![Page 1: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/1.jpg)
Descriptional complexity of automata andlanguages
Giovanni Pighizzini
Dipartimento di Informatica e ComunicazioneUniversità degli Studi di Milano
ITALY
Milano – December 10th, 2008
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 2: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/2.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 3: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/3.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 4: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/4.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 5: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/5.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 6: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/6.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 7: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/7.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 8: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/8.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 9: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/9.jpg)
Outline of the talk
1 Introduction
2 Regular languages and finite automata
3 The problem of Sakoda and Sipser
4 Unary automata
5 Regular languages vs context-free grammars andpushdown automata
6 State complexity of language operations
7 State complexity of the complementation
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 10: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/10.jpg)
Descriptional complexity
GivenC, a class of languagesS, a formal system (e.g., class of devices, class ofgrammars,..) able to represent all the languages in C
What is the size of the representations of the languagesin C by the system S?
Usually, descriptional complexity compares different descriptionfor a same class of languages:
given S ′, another formal system able to represent all thelanguages in C
What is the size of the representations of the languagesin C by the system S ′, with respect to the size of their
representations by the system S?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 11: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/11.jpg)
Descriptional complexity
GivenC, a class of languagesS, a formal system (e.g., class of devices, class ofgrammars,..) able to represent all the languages in C
What is the size of the representations of the languagesin C by the system S?
Usually, descriptional complexity compares different descriptionfor a same class of languages:
given S ′, another formal system able to represent all thelanguages in C
What is the size of the representations of the languagesin C by the system S ′, with respect to the size of their
representations by the system S?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 12: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/12.jpg)
Descriptional complexity
GivenC, a class of languagesS, a formal system (e.g., class of devices, class ofgrammars,..) able to represent all the languages in C
What is the size of the representations of the languagesin C by the system S?
Usually, descriptional complexity compares different descriptionfor a same class of languages:
given S ′, another formal system able to represent all thelanguages in C
What is the size of the representations of the languagesin C by the system S ′, with respect to the size of their
representations by the system S?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 13: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/13.jpg)
Descriptional complexity
GivenC, a class of languagesS, a formal system (e.g., class of devices, class ofgrammars,..) able to represent all the languages in C
What is the size of the representations of the languagesin C by the system S?
Usually, descriptional complexity compares different descriptionfor a same class of languages:
given S ′, another formal system able to represent all thelanguages in C
What is the size of the representations of the languagesin C by the system S ′, with respect to the size of their
representations by the system S?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 14: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/14.jpg)
Descriptional complexity
GivenC, a class of languagesS, a formal system (e.g., class of devices, class ofgrammars,..) able to represent all the languages in C
What is the size of the representations of the languagesin C by the system S?
Usually, descriptional complexity compares different descriptionfor a same class of languages:
given S ′, another formal system able to represent all thelanguages in C
What is the size of the representations of the languagesin C by the system S ′, with respect to the size of their
representations by the system S?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 15: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/15.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 16: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/16.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Base version:
One-way determistic finite automata (1dfa)one-way input tapedeterministic
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 17: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/17.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinismtwo-way input head motionalternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 18: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/18.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinismtwo-way input head motionalternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 19: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/19.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinism
one-way nondetermistic finite automata (1nfa)
two-way input head motionalternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 20: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/20.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinism (1nfa)two-way input head motionalternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 21: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/21.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinism (1nfa)two-way input head motion
two-way determistic finite automata (2dfa)two-way nondetermistic finite automata (2nfa)
alternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 22: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/22.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinism (1nfa)two-way input head motion
two-way determistic finite automata (2dfa)two-way nondetermistic finite automata (2nfa)
alternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 23: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/23.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinism (1nfa)two-way input head motion (2dfa, 2nfa)alternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 24: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/24.jpg)
Descriptional complexity
Classical example: finite state automata
i n p u t. . .
6-
Some possibile variants introducing:non determinism (1nfa)two-way input head motion (2dfa, 2nfa)alternation...
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 25: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/25.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 26: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/26.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 27: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/27.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 28: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/28.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 29: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/29.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 30: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/30.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 31: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/31.jpg)
Finite state automata: different variants
1dfa, 1nfa, 2dfa, 2nfa, ...Formal language point of view:
What about the power of these models?All of them characterize the class of regular languages...
Descriptional complexity point of view:What about the size of their descriptions?
...however some of them are more succinct.
Example
Each n-state 1nfa can be simulated by a 1dfa with 2n
states (subset construction) [Rabin and Scott ’59], and:For each integer n ≥ 1 there is a language which isaccepted by a n-state 1nfa which requires 2n states to beaccepted by a 1dfa [Meyer and Fischer ’71].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 32: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/32.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 33: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/33.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 34: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/34.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 35: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/35.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 36: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/36.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 37: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/37.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 38: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/38.jpg)
Descriptional complexity measures for finite automa
Deterministic automata (dfa):number of states.
Nondeterministic automata (nfa):number of states, ornumber of transitions (more precise).
Remarks:
Each nfa with n states has O(n2) transitions.Many results have been obtained for the measure “numberof the states”.The measure “number of transitions” was recentlyinvestigated by some authors,however it was already proposed as a more realisticmeasure in1971 by Meyer and Fischer.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 39: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/39.jpg)
Finite state automata
Costs of the optimal simulations by 1dfa (in terms of states):
1dfa
1nfa 2dfa 2nfa
@@
@@
@@
@R ?
2n O(2log n) O(2n2)
-? ?
[Rabin and Scott ’59, Shepardson ’59,Meyer and Fischer ’71,...]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 40: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/40.jpg)
Finite state automata
Costs of the optimal simulations by 1dfa (in terms of states):
1dfa
1nfa 2dfa 2nfa
@@
@@
@@
@R ?
2n O(2log n) O(2n2)
-? ?
How much two-way motion can be useful inorder to eliminate the nondeterminism?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 41: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/41.jpg)
Finite state automata
Costs of the optimal simulations by 1dfa (in terms of states):
1dfa
1nfa 2dfa 2nfa
@@
@@
@@
@R ?
2n O(2log n) O(2n2)
-? ?
Problem ([Sakoda and Sipser 1978])Find the costs, in terms of states, of the optimal simulations of
1nfa by 2dfa2nfa by 2dfa
Conjecture: these costs are exponentialGiovanni Pighizzini Descriptional complexity of automata and languages
![Page 42: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/42.jpg)
Finite state automata
Costs of the optimal simulations by 1dfa (in terms of states):
1dfa
1nfa 2dfa 2nfa
@@
@@
@@
@R ?
2n O(2log n) O(2n2)
-?
?
Problem ([Sakoda and Sipser 1978])Find the costs, in terms of states, of the optimal simulations of
1nfa by 2dfa2nfa by 2dfa
Conjecture: these costs are exponentialGiovanni Pighizzini Descriptional complexity of automata and languages
![Page 43: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/43.jpg)
Finite state automata
Costs of the optimal simulations by 1dfa (in terms of states):
1dfa
1nfa 2dfa 2nfa
@@
@@
@@
@R ?
2n O(2log n) O(2n2)
-? ?
Problem ([Sakoda and Sipser 1978])Find the costs, in terms of states, of the optimal simulations of
1nfa by 2dfa2nfa by 2dfa
Conjecture: these costs are exponentialGiovanni Pighizzini Descriptional complexity of automata and languages
![Page 44: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/44.jpg)
Finite state automata
Costs of the optimal simulations by 1dfa (in terms of states):
1dfa
1nfa 2dfa 2nfa
@@
@@
@@
@R ?
2n O(2log n) O(2n2)
-? ?
Problem ([Sakoda and Sipser 1978])Find the costs, in terms of states, of the optimal simulations of
1nfa by 2dfa2nfa by 2dfa
Conjecture: these costs are exponentialGiovanni Pighizzini Descriptional complexity of automata and languages
![Page 45: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/45.jpg)
Sakoda and Sipser question: complete languages
Theorem (Complete languages for 1nfa vs 2dfa)
There exists a sequence of languages < B1, B2, . . . , Bn, . . . >s.t. for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andamong all languages accepted by n-state 1nfa, Bn requiresthe largest 2dfa.
Remark: the second condition implies thatthe simulation of 1nfa by 2dfa is polynomial iff each Bn isaccepted by a 2dfa with a polynomial (in n) number of states.
In a similar way:
Theorem (Complete languages for 2nfa vs 2dfa)
There exists a sequence of languages < C1, C2, . . . , Cn, . . . >complete for the reduction of 2nfa to 2dfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 46: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/46.jpg)
Sakoda and Sipser question: complete languages
Theorem (Complete languages for 1nfa vs 2dfa)
There exists a sequence of languages < B1, B2, . . . , Bn, . . . >s.t. for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andamong all languages accepted by n-state 1nfa, Bn requiresthe largest 2dfa.
Remark: the second condition implies thatthe simulation of 1nfa by 2dfa is polynomial iff each Bn isaccepted by a 2dfa with a polynomial (in n) number of states.
In a similar way:
Theorem (Complete languages for 2nfa vs 2dfa)
There exists a sequence of languages < C1, C2, . . . , Cn, . . . >complete for the reduction of 2nfa to 2dfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 47: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/47.jpg)
Sakoda and Sipser question: complete languages
Theorem (Complete languages for 1nfa vs 2dfa)
There exists a sequence of languages < B1, B2, . . . , Bn, . . . >s.t. for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andamong all languages accepted by n-state 1nfa, Bn requiresthe largest 2dfa.
Remark: the second condition implies thatthe simulation of 1nfa by 2dfa is polynomial iff each Bn isaccepted by a 2dfa with a polynomial (in n) number of states.
In a similar way:
Theorem (Complete languages for 2nfa vs 2dfa)
There exists a sequence of languages < C1, C2, . . . , Cn, . . . >complete for the reduction of 2nfa to 2dfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 48: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/48.jpg)
Sakoda and Sipser question: complete languages
Theorem (Complete languages for 1nfa vs 2dfa)
There exists a sequence of languages < B1, B2, . . . , Bn, . . . >s.t. for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andamong all languages accepted by n-state 1nfa, Bn requiresthe largest 2dfa.
Remark: the second condition implies thatthe simulation of 1nfa by 2dfa is polynomial iff each Bn isaccepted by a 2dfa with a polynomial (in n) number of states.
In a similar way:
Theorem (Complete languages for 2nfa vs 2dfa)
There exists a sequence of languages < C1, C2, . . . , Cn, . . . >complete for the reduction of 2nfa to 2dfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 49: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/49.jpg)
Sakoda and Sipser question: complete languages
Theorem (Complete languages for 1nfa vs 2dfa)
There exists a sequence of languages < B1, B2, . . . , Bn, . . . >s.t. for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andamong all languages accepted by n-state 1nfa, Bn requiresthe largest 2dfa.
Remark: the second condition implies thatthe simulation of 1nfa by 2dfa is polynomial iff each Bn isaccepted by a 2dfa with a polynomial (in n) number of states.
In a similar way:
Theorem (Complete languages for 2nfa vs 2dfa)
There exists a sequence of languages < C1, C2, . . . , Cn, . . . >complete for the reduction of 2nfa to 2dfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 50: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/50.jpg)
Sakoda and Sipser question: lower bounds
Polynomial lower bounds have been proved for the cost c(n) ofsimulation of 1nfa by 2dfa. In particular:
c(n) ∈ Ω( n2
log n ) [Berman and Lingas 1977]
c(n) ∈ Ω(n2) [Chrobak 1986]
Exponential lower bounds have been proved, if the resultingmachines are required to satisfy some special conditions. e.g.,
sweeping automata [Sipser 1980]
oblivious automata [Hromkovic and Schnitger 2003]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 51: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/51.jpg)
Sakoda and Sipser question: lower bounds
Polynomial lower bounds have been proved for the cost c(n) ofsimulation of 1nfa by 2dfa. In particular:
c(n) ∈ Ω( n2
log n ) [Berman and Lingas 1977]
c(n) ∈ Ω(n2) [Chrobak 1986]
Exponential lower bounds have been proved, if the resultingmachines are required to satisfy some special conditions. e.g.,
sweeping automata [Sipser 1980]
oblivious automata [Hromkovic and Schnitger 2003]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 52: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/52.jpg)
Sakoda and Sipser question: lower bounds
Polynomial lower bounds have been proved for the cost c(n) ofsimulation of 1nfa by 2dfa. In particular:
c(n) ∈ Ω( n2
log n ) [Berman and Lingas 1977]
c(n) ∈ Ω(n2) [Chrobak 1986]
Exponential lower bounds have been proved, if the resultingmachines are required to satisfy some special conditions. e.g.,
sweeping automata [Sipser 1980]
oblivious automata [Hromkovic and Schnitger 2003]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 53: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/53.jpg)
Sakoda and Sipser question: lower bounds
Polynomial lower bounds have been proved for the cost c(n) ofsimulation of 1nfa by 2dfa. In particular:
c(n) ∈ Ω( n2
log n ) [Berman and Lingas 1977]
c(n) ∈ Ω(n2) [Chrobak 1986]
Exponential lower bounds have been proved, if the resultingmachines are required to satisfy some special conditions. e.g.,
sweeping automata [Sipser 1980]
oblivious automata [Hromkovic and Schnitger 2003]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 54: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/54.jpg)
Sakoda and Sipser question: lower bounds
Polynomial lower bounds have been proved for the cost c(n) ofsimulation of 1nfa by 2dfa. In particular:
c(n) ∈ Ω( n2
log n ) [Berman and Lingas 1977]
c(n) ∈ Ω(n2) [Chrobak 1986]
Exponential lower bounds have been proved, if the resultingmachines are required to satisfy some special conditions. e.g.,
sweeping automata [Sipser 1980]
oblivious automata [Hromkovic and Schnitger 2003]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 55: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/55.jpg)
Sakoda and Sipser question: lower bounds
Polynomial lower bounds have been proved for the cost c(n) ofsimulation of 1nfa by 2dfa. In particular:
c(n) ∈ Ω( n2
log n ) [Berman and Lingas 1977]
c(n) ∈ Ω(n2) [Chrobak 1986]
Exponential lower bounds have been proved, if the resultingmachines are required to satisfy some special conditions. e.g.,
sweeping automata [Sipser 1980]
oblivious automata [Hromkovic and Schnitger 2003]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 56: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/56.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 57: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/57.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 58: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/58.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 59: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/59.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 60: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/60.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 61: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/61.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 62: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/62.jpg)
Sweeping automata
Theorem ([Sipser 1980])There exists a family of languages < B1, B2, . . . , Bn, . . . > s.t.for each integer n ≥ 1:
Bn is accepted by a 1nfa with n states, andBn cannot be accepted by any sweeping automaton withless than 2n states.
However, 2dfa can be exponentially more succinct thansweeping automata:
Theorem ([Berman 1981, Micali 1981])There exists a family of languages < A1, A2, . . . , An, . . . > s.t.for each integer n ≥ 1:
An is accepted by a 2dfa with n states, andAn cannot be accepted by any sweeping automaton withless than 2n − 1 states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 63: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/63.jpg)
Sakoda&Sipser question
Current knowledge
Upper bounds
general case
unary case
1nfa → 2dfa 2nfa → 2dfa
exponential exponential
O(n2) [1] nO(log n) [2]optimal
[1: Chrobak 1986][2: Geffert, Mereghetti, Pighizzini 2003]
Lower boundsFor all the cases, the best known lower bound is Ω(n2) [1]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 64: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/64.jpg)
Sakoda&Sipser question
Current knowledge
Upper bounds
general case
unary case
1nfa → 2dfa 2nfa → 2dfa
exponential exponential
O(n2) [1] nO(log n) [2]optimal
[1: Chrobak 1986][2: Geffert, Mereghetti, Pighizzini 2003]
Lower boundsFor all the cases, the best known lower bound is Ω(n2) [1]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 65: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/65.jpg)
Sakoda&Sipser question
Current knowledge
Upper bounds
general case
unary case
1nfa → 2dfa 2nfa → 2dfa
exponential exponential
O(n2) [1] nO(log n) [2]optimal
[1: Chrobak 1986][2: Geffert, Mereghetti, Pighizzini 2003]
Lower boundsFor all the cases, the best known lower bound is Ω(n2) [1]
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 66: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/66.jpg)
Unary automata: 1dfa
Input alphabet Σ = a
- m - m - m - mj m
mm
TheoremL ⊆ a∗ is regular iff ∃µ ≥ 0, λ ≥ 1 s.t.
∀n ≥ µ : an ∈ L iff an+λ ∈ L.
Special case µ = 0: the language is periodic or cyclic.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 67: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/67.jpg)
Unary automata: 1dfa
Input alphabet Σ = a
- m - m - m - mj m
mm
TheoremL ⊆ a∗ is regular iff ∃µ ≥ 0, λ ≥ 1 s.t.
∀n ≥ µ : an ∈ L iff an+λ ∈ L.
Special case µ = 0: the language is periodic or cyclic.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 68: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/68.jpg)
Unary automata: 1dfa
Input alphabet Σ = a
- m - m - m - mj m
mm
TheoremL ⊆ a∗ is regular iff ∃µ ≥ 0, λ ≥ 1 s.t.
∀n ≥ µ : an ∈ L iff an+λ ∈ L.
Special case µ = 0: the language is periodic or cyclic.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 69: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/69.jpg)
Unary automata: 1dfa
Input alphabet Σ = a
- m - m - m︸ ︷︷ ︸µ
- mj m
mm
︸ ︷︷ ︸λ
TheoremL ⊆ a∗ is regular iff ∃µ ≥ 0, λ ≥ 1 s.t.
∀n ≥ µ : an ∈ L iff an+λ ∈ L.
Special case µ = 0: the language is periodic or cyclic.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 70: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/70.jpg)
Unary automata: 1dfa
Input alphabet Σ = a
- m - m - m︸ ︷︷ ︸µ
- mj m
mm
︸ ︷︷ ︸λ
TheoremL ⊆ a∗ is regular iff ∃µ ≥ 0, λ ≥ 1 s.t.
∀n ≥ µ : an ∈ L iff an+λ ∈ L.
Special case µ = 0: the language is periodic or cyclic.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 71: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/71.jpg)
Unary automata: 1dfa
Input alphabet Σ = a
- m - m - m︸ ︷︷ ︸µ
- mj m
mm
︸ ︷︷ ︸λ
TheoremL ⊆ a∗ is regular iff ∃µ ≥ 0, λ ≥ 1 s.t.
∀n ≥ µ : an ∈ L iff an+λ ∈ L.
Special case µ = 0: the language is periodic or cyclic.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 72: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/72.jpg)
Unary automata: 1nfa
The transition graph describing the automaton can have awhatever structure, however...
...we can restrict to 1nfa with the following form (Chrobaknormal form):
an initial patha nondeterministic choicea set of cycles
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 73: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/73.jpg)
Unary automata: 1nfa
The transition graph describing the automaton can have awhatever structure, however...
...we can restrict to 1nfa with the following form (Chrobaknormal form):
an initial patha nondeterministic choicea set of cycles
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 74: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/74.jpg)
Unary automata: 1nfa
The transition graph describing the automaton can have awhatever structure, however...
...we can restrict to 1nfa with the following form (Chrobaknormal form):
an initial patha nondeterministic choicea set of cycles
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 75: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/75.jpg)
Unary automata: 1nfa
The transition graph describing the automaton can have awhatever structure, however...
...we can restrict to 1nfa with the following form (Chrobaknormal form):
an initial patha nondeterministic choicea set of cycles
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 76: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/76.jpg)
Unary automata: 1nfa
The transition graph describing the automaton can have awhatever structure, however...
...we can restrict to 1nfa with the following form (Chrobaknormal form):
an initial patha nondeterministic choicea set of cycles
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 77: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/77.jpg)
Unary automata: 1nfa
The transition graph describing the automaton can have awhatever structure, however...
...we can restrict to 1nfa with the following form (Chrobaknormal form):
an initial patha nondeterministic choicea set of cycles
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 78: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/78.jpg)
Unary automata: Chrobak normal form
TheoremFor any unary n-state 1nfa there exists an equivalent 1nfa inChrobak normal form s.t.
there are O(n2) states on the initial paththe total number of states on the cycles is at most n
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 79: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/79.jpg)
Unary automata: Chrobak normal form
TheoremFor any unary n-state 1nfa there exists an equivalent 1nfa inChrobak normal form s.t.
there are O(n2) states on the initial paththe total number of states on the cycles is at most n
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 80: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/80.jpg)
Unary automata: Chrobak normal form
TheoremFor any unary n-state 1nfa there exists an equivalent 1nfa inChrobak normal form s.t.
there are O(n2) states on the initial paththe total number of states on the cycles is at most n
- m a- km a- m
a
@@@R
a
m mk-a
a
mk mm
a
?a
a
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 81: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/81.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 82: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/82.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 83: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/83.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 84: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/84.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 85: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/85.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 86: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/86.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 87: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/87.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 88: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/88.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 89: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/89.jpg)
Unary nondeterministic automata
How to eliminate the nondeterminism from a unary 1nfa?1 Convert the given 1nfa to
the Chrobak normal form:
- m a- m a- m
a
@@@R
a
0 1 2
m m0 1-a
a
m mm
0 2
1
a
?a
a
kk
k
2 Copy the initial path3 Replace the set of cycles
with a unique cycle:
- m a- m a- mk0 1 2 a- ma
km a-k m
a@@R ma
kmama@
@I k00
11 02
10
0112
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 90: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/90.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 91: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/91.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 92: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/92.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 93: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/93.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 94: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/94.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 95: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/95.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 96: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/96.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 97: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/97.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 98: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/98.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 99: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/99.jpg)
Simulation of unary 1nfa by 1dfa
Given an n-state 1nfa:Convert it in Chrobak normal form:
Initial path of O(n2) statesCycles of lenghts λ1, . . . , λk , with λ1 + . . . + λk ≤ n
The cycles are replaced by a unique cycle of lengthlcm(λ1, . . . , λk )
Hence:The number of states in the resulting cycle is bouned byF (n) = maxlcm(x1, . . . , xk ) | x1 + . . . + xk = n
and:F (n) = eO(
√n log n) [Landau 1903]
Theorem ([Chrobak 1986])Each unary n-state 1nfa can be simulated by a 1dfa witheO(
√n log n) states.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 100: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/100.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 101: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/101.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 102: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/102.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 103: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/103.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
6
eO(√
n log n)
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 104: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/104.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
6
eO(√
n log n)
@@
@@
@@
@@I
eO(√
n log n)
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 105: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/105.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
6
eO(√
n log n)
@@
@@
@@
@@I
eO(√
n log n)
6
eO(√
n log n)
# ?
eO(√
n log n)
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 106: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/106.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
6
eO(√
n log n)
@@
@@
@@
@@I
eO(√
n log n)
6
eO(√
n log n)
# ?
eO(√
n log n)
!6 n2
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 107: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/107.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
6
eO(√
n log n)
@@
@@
@@
@@I
eO(√
n log n)
6
eO(√
n log n)
# ?
eO(√
n log n)
!6 n2
?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 108: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/108.jpg)
Unary automata
The costs of the optimal simulations between automataare different in the unary and in the general case!
Unary case: [Chrobak 1986, Mereghetti and Pighizzini 2001]
2dfa
1dfa 1nfa
2nfa
eO(
√n log n)
6
eO(√
n log n)
@@
@@
@@
@@I
eO(√
n log n)
6
eO(√
n log n)
# ?
eO(√
n log n)
!6 n2
?
-n
@@
@@
@@
@@R
n
?
n
?
n
6
eO(√
n log n)
-n?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 109: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/109.jpg)
Descriptional complexity of regular languages
Different variant of finite automata characterize regularlanguages.However, we can describe regular languages using morepowerful formalisms or devices, as context-free grammarsand pushdown automata.
What about the sizes of cfg’s or pda’s describing regularlanguages vs the sizes of finite automata?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 110: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/110.jpg)
Descriptional complexity of regular languages
Different variant of finite automata characterize regularlanguages.However, we can describe regular languages using morepowerful formalisms or devices, as context-free grammarsand pushdown automata.
What about the sizes of cfg’s or pda’s describing regularlanguages vs the sizes of finite automata?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 111: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/111.jpg)
Descriptional complexity of regular languages
Different variant of finite automata characterize regularlanguages.However, we can describe regular languages using morepowerful formalisms or devices, as context-free grammarsand pushdown automata.
What about the sizes of cfg’s or pda’s describing regularlanguages vs the sizes of finite automata?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 112: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/112.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 113: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/113.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 114: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/114.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 115: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/115.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 116: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/116.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 117: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/117.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 118: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/118.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 119: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/119.jpg)
Descriptional complexity measures
Context-free grammars:number of variables?
For n ≥ 1, consider the language Ln = (an)∗:Ln requires n states to be accepted by dfa or nfaLn is generated by the grammar with one variable S and theproductions
S → an S → anS S → ε
Thus, the number of variables cannot be a descriptionalcomplexity measure for context-free grammars.However, for grammars in Chomsky Normal Form thenumber of variables is a “reasonable” measure ofcomplexity [Gruska, 1973].
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 120: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/120.jpg)
Descriptional complexity measures
Pushdown automata:W.l.o.g., we can consider only pda’s s.t. push operationsadd exactly one symbol on the pushdown store.
The total size of the description of a pda satisfying thisrestriction is a polynomial function of two parameters:
the number of the statesthe cardinality of the pushdown alphabet.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 121: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/121.jpg)
Descriptional complexity measures
Pushdown automata:W.l.o.g., we can consider only pda’s s.t. push operationsadd exactly one symbol on the pushdown store.
The total size of the description of a pda satisfying thisrestriction is a polynomial function of two parameters:
the number of the statesthe cardinality of the pushdown alphabet.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 122: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/122.jpg)
Context-free vs regular: descriptional complexity
Given a context-free grammar (or a pushdown automaton) ofsize n, generating a regular language, how much is big anequivalent finite automaton, wrt n ?
Theorem ([Meyer and Fischer, 1971])For any recursive function f and arbitrarily large integers n,there exists a cfg G of size n generating a regular language L,s.t. any dfa accepting L must have at least f (n) states.
As a consequence, the trade-off between context-grammarsand finite automata is not recursive.However... The witness language L is defined over a binaryalphabet.
What about languages over a one letteralphabet?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 123: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/123.jpg)
Context-free vs regular: descriptional complexity
Given a context-free grammar (or a pushdown automaton) ofsize n, generating a regular language, how much is big anequivalent finite automaton, wrt n ?
Theorem ([Meyer and Fischer, 1971])For any recursive function f and arbitrarily large integers n,there exists a cfg G of size n generating a regular language L,s.t. any dfa accepting L must have at least f (n) states.
As a consequence, the trade-off between context-grammarsand finite automata is not recursive.However... The witness language L is defined over a binaryalphabet.
What about languages over a one letteralphabet?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 124: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/124.jpg)
Context-free vs regular: descriptional complexity
Given a context-free grammar (or a pushdown automaton) ofsize n, generating a regular language, how much is big anequivalent finite automaton, wrt n ?
Theorem ([Meyer and Fischer, 1971])For any recursive function f and arbitrarily large integers n,there exists a cfg G of size n generating a regular language L,s.t. any dfa accepting L must have at least f (n) states.
As a consequence, the trade-off between context-grammarsand finite automata is not recursive.However... The witness language L is defined over a binaryalphabet.
What about languages over a one letteralphabet?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 125: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/125.jpg)
Context-free vs regular: descriptional complexity
Given a context-free grammar (or a pushdown automaton) ofsize n, generating a regular language, how much is big anequivalent finite automaton, wrt n ?
Theorem ([Meyer and Fischer, 1971])For any recursive function f and arbitrarily large integers n,there exists a cfg G of size n generating a regular language L,s.t. any dfa accepting L must have at least f (n) states.
As a consequence, the trade-off between context-grammarsand finite automata is not recursive.However... The witness language L is defined over a binaryalphabet.
What about languages over a one letteralphabet?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 126: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/126.jpg)
Context-free vs regular: descriptional complexity
Given a context-free grammar (or a pushdown automaton) ofsize n, generating a regular language, how much is big anequivalent finite automaton, wrt n ?
Theorem ([Meyer and Fischer, 1971])For any recursive function f and arbitrarily large integers n,there exists a cfg G of size n generating a regular language L,s.t. any dfa accepting L must have at least f (n) states.
As a consequence, the trade-off between context-grammarsand finite automata is not recursive.However... The witness language L is defined over a binaryalphabet.
What about languages over a one letteralphabet?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 127: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/127.jpg)
Unary languages
Σ = a
Theorem ([Ginsurg and Rice, 1962])Every unary context-free language is regular.
Hence the classes of unary regular languages and unarycontext-free languages coincide!
ProblemStudy the equivalence between unary context-free and regularlanguages from the descriptional complexity point of view.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 128: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/128.jpg)
Unary languages
Σ = a
Theorem ([Ginsurg and Rice, 1962])Every unary context-free language is regular.
Hence the classes of unary regular languages and unarycontext-free languages coincide!
ProblemStudy the equivalence between unary context-free and regularlanguages from the descriptional complexity point of view.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 129: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/129.jpg)
Unary cfg’s vs finite automata
The following bounds have been proved in[Pighizzini, Shallit, Wang, 2002]:
Given a unary cfg in Chomsky normal form with h variables,there exist:
an equivalent 1nfa with at most 22h−1 + 1 statesan equivalent 1dfa with at most 2h2
states
These bounds are tight, namely, matching lower bound havebeen discovered.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 130: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/130.jpg)
Unary cfg’s vs finite automata
The following bounds have been proved in[Pighizzini, Shallit, Wang, 2002]:
Given a unary cfg in Chomsky normal form with h variables,there exist:
an equivalent 1nfa with at most 22h−1 + 1 statesan equivalent 1dfa with at most 2h2
states
These bounds are tight, namely, matching lower bound havebeen discovered.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 131: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/131.jpg)
Unary cfg’s vs finite automata
The following bounds have been proved in[Pighizzini, Shallit, Wang, 2002]:
Given a unary cfg in Chomsky normal form with h variables,there exist:
an equivalent 1nfa with at most 22h−1 + 1 statesan equivalent 1dfa with at most 2h2
states
These bounds are tight, namely, matching lower bound havebeen discovered.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 132: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/132.jpg)
Unary cfg’s vs finite automata
The following bounds have been proved in[Pighizzini, Shallit, Wang, 2002]:
Given a unary cfg in Chomsky normal form with h variables,there exist:
an equivalent 1nfa with at most 22h−1 + 1 statesan equivalent 1dfa with at most 2h2
states
These bounds are tight, namely, matching lower bound havebeen discovered.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 133: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/133.jpg)
Extension to bounded languages
Bounded languages:Subsets of w∗
1 w∗2 . . . w∗
n , for given words w1, . . . , wn(letter bounded if w1, . . . , wn ∈ Σ).
The class bounded regular languages is properly includedin that of bounded cfl’s, e.g., anbn | n ≥ 0.Bounded cfl’s can be accepted by finite turn pda’s.
Theorem ([Malcher, Pighizzini, 2007])Each bounded context-free language generated by a cfg with hvariables in Chomsky normal form is accepted by a finite-turnpda with 2h states and O(1) stack symbols.
Even this upper bound is tight!
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 134: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/134.jpg)
Extension to bounded languages
Bounded languages:Subsets of w∗
1 w∗2 . . . w∗
n , for given words w1, . . . , wn(letter bounded if w1, . . . , wn ∈ Σ).
The class bounded regular languages is properly includedin that of bounded cfl’s, e.g., anbn | n ≥ 0.Bounded cfl’s can be accepted by finite turn pda’s.
Theorem ([Malcher, Pighizzini, 2007])Each bounded context-free language generated by a cfg with hvariables in Chomsky normal form is accepted by a finite-turnpda with 2h states and O(1) stack symbols.
Even this upper bound is tight!
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 135: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/135.jpg)
Extension to bounded languages
Bounded languages:Subsets of w∗
1 w∗2 . . . w∗
n , for given words w1, . . . , wn(letter bounded if w1, . . . , wn ∈ Σ).
The class bounded regular languages is properly includedin that of bounded cfl’s, e.g., anbn | n ≥ 0.Bounded cfl’s can be accepted by finite turn pda’s.
Theorem ([Malcher, Pighizzini, 2007])Each bounded context-free language generated by a cfg with hvariables in Chomsky normal form is accepted by a finite-turnpda with 2h states and O(1) stack symbols.
Even this upper bound is tight!
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 136: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/136.jpg)
Extension to bounded languages
Bounded languages:Subsets of w∗
1 w∗2 . . . w∗
n , for given words w1, . . . , wn(letter bounded if w1, . . . , wn ∈ Σ).
The class bounded regular languages is properly includedin that of bounded cfl’s, e.g., anbn | n ≥ 0.Bounded cfl’s can be accepted by finite turn pda’s.
Theorem ([Malcher, Pighizzini, 2007])Each bounded context-free language generated by a cfg with hvariables in Chomsky normal form is accepted by a finite-turnpda with 2h states and O(1) stack symbols.
Even this upper bound is tight!
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 137: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/137.jpg)
Extension to bounded languages
Bounded languages:Subsets of w∗
1 w∗2 . . . w∗
n , for given words w1, . . . , wn(letter bounded if w1, . . . , wn ∈ Σ).
The class bounded regular languages is properly includedin that of bounded cfl’s, e.g., anbn | n ≥ 0.Bounded cfl’s can be accepted by finite turn pda’s.
Theorem ([Malcher, Pighizzini, 2007])Each bounded context-free language generated by a cfg with hvariables in Chomsky normal form is accepted by a finite-turnpda with 2h states and O(1) stack symbols.
Even this upper bound is tight!
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 138: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/138.jpg)
Unary dpda’s vs finite automata
Let M be a unary pda with n states and m stack symbols, s.t.each push adds exactly one symbol.
Using the simulation of unary cfg’s by finite automata and thestandard transformation of pda’s into cfg’s, it can be shown thatM can be simulated by a 1dfa with with 2O(n4m2) states.
What about the deterministic case?
Theorem ([Pighizzini, 2008])If M is deterministic then it can be simulated by a 1dfawith 2O(nm) states.Furthermore, such a simulation is tight.Its cost cannot be reduced even if we simulate M by a 2nfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 139: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/139.jpg)
Unary dpda’s vs finite automata
Let M be a unary pda with n states and m stack symbols, s.t.each push adds exactly one symbol.
Using the simulation of unary cfg’s by finite automata and thestandard transformation of pda’s into cfg’s, it can be shown thatM can be simulated by a 1dfa with with 2O(n4m2) states.
What about the deterministic case?
Theorem ([Pighizzini, 2008])If M is deterministic then it can be simulated by a 1dfawith 2O(nm) states.Furthermore, such a simulation is tight.Its cost cannot be reduced even if we simulate M by a 2nfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 140: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/140.jpg)
Unary dpda’s vs finite automata
Let M be a unary pda with n states and m stack symbols, s.t.each push adds exactly one symbol.
Using the simulation of unary cfg’s by finite automata and thestandard transformation of pda’s into cfg’s, it can be shown thatM can be simulated by a 1dfa with with 2O(n4m2) states.
What about the deterministic case?
Theorem ([Pighizzini, 2008])If M is deterministic then it can be simulated by a 1dfawith 2O(nm) states.Furthermore, such a simulation is tight.Its cost cannot be reduced even if we simulate M by a 2nfa.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 141: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/141.jpg)
Languages with “complex” dpda’s
Unary dpda’s can be exponentially more succinct than dfa’s.Does this is true for each unary regular language?
ProblemFor m ≥ 0, let Lm ⊆ a∗ be a language accepted by a dfa with2m states.Does there exists an equivalent dpda with O(m) states?
The answer to this question is negative: [Pighizzini, 2008]
For each m > 0 there exists a language Lm ⊆ a∗ s.t.:Lm is accepted by a dfa with 2m states.The size of any dpda accepting Lm is at least d 2m
m2 , for aconstant d .
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 142: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/142.jpg)
Languages with “complex” dpda’s
Unary dpda’s can be exponentially more succinct than dfa’s.Does this is true for each unary regular language?
ProblemFor m ≥ 0, let Lm ⊆ a∗ be a language accepted by a dfa with2m states.Does there exists an equivalent dpda with O(m) states?
The answer to this question is negative: [Pighizzini, 2008]
For each m > 0 there exists a language Lm ⊆ a∗ s.t.:Lm is accepted by a dfa with 2m states.The size of any dpda accepting Lm is at least d 2m
m2 , for aconstant d .
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 143: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/143.jpg)
Languages with “complex” dpda’s
Unary dpda’s can be exponentially more succinct than dfa’s.Does this is true for each unary regular language?
ProblemFor m ≥ 0, let Lm ⊆ a∗ be a language accepted by a dfa with2m states.Does there exists an equivalent dpda with O(m) states?
The answer to this question is negative: [Pighizzini, 2008]
For each m > 0 there exists a language Lm ⊆ a∗ s.t.:Lm is accepted by a dfa with 2m states.The size of any dpda accepting Lm is at least d 2m
m2 , for aconstant d .
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 144: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/144.jpg)
State complexity of language operations
Let L1 and L2 two regular languages accepted by twoautomata with s1 and s2 states, and op a binary operationpreserving the regularity.
State a tight upper bound fop(s1, s2) for the number of thestates of an automaton accepting the language L1op L2.
A same question can be formulated for other kinds ofoperations.
These problems have been extensively studied in the case ofone-way deterministic and nondeterministic automata for“classical” regular operations.
There are interesting results and problems for the complementof languages in the case of nondeterministic machines.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 145: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/145.jpg)
State complexity of language operations
Let L1 and L2 two regular languages accepted by twoautomata with s1 and s2 states, and op a binary operationpreserving the regularity.
State a tight upper bound fop(s1, s2) for the number of thestates of an automaton accepting the language L1op L2.
A same question can be formulated for other kinds ofoperations.
These problems have been extensively studied in the case ofone-way deterministic and nondeterministic automata for“classical” regular operations.
There are interesting results and problems for the complementof languages in the case of nondeterministic machines.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 146: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/146.jpg)
State complexity of language operations
Let L1 and L2 two regular languages accepted by twoautomata with s1 and s2 states, and op a binary operationpreserving the regularity.
State a tight upper bound fop(s1, s2) for the number of thestates of an automaton accepting the language L1op L2.
A same question can be formulated for other kinds ofoperations.
These problems have been extensively studied in the case ofone-way deterministic and nondeterministic automata for“classical” regular operations.
There are interesting results and problems for the complementof languages in the case of nondeterministic machines.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 147: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/147.jpg)
State complexity of language operations
Let L1 and L2 two regular languages accepted by twoautomata with s1 and s2 states, and op a binary operationpreserving the regularity.
State a tight upper bound fop(s1, s2) for the number of thestates of an automaton accepting the language L1op L2.
A same question can be formulated for other kinds ofoperations.
These problems have been extensively studied in the case ofone-way deterministic and nondeterministic automata for“classical” regular operations.
There are interesting results and problems for the complementof languages in the case of nondeterministic machines.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 148: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/148.jpg)
State complexity of language operations
Let L1 and L2 two regular languages accepted by twoautomata with s1 and s2 states, and op a binary operationpreserving the regularity.
State a tight upper bound fop(s1, s2) for the number of thestates of an automaton accepting the language L1op L2.
A same question can be formulated for other kinds ofoperations.
These problems have been extensively studied in the case ofone-way deterministic and nondeterministic automata for“classical” regular operations.
There are interesting results and problems for the complementof languages in the case of nondeterministic machines.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 149: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/149.jpg)
Complementation of two-way automata
ProblemGiven a two-way automaton with n states accepting a languageL, find the cost in term of states, of an automaton accepting thecomplement of L.
Deterministic case:The cost is 4n, for any input alphabet.[Geffert, Mereghetti, Pighizzini 2007](note that 2dfa’s can have infinite computations)Nondeterministic case:The cost is polynomial for a unary alphabet.[Geffert, Mereghetti, Pighizzini 2007]
What about the complementation of 2nfa over nonunaryalphabets?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 150: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/150.jpg)
Complementation of two-way automata
ProblemGiven a two-way automaton with n states accepting a languageL, find the cost in term of states, of an automaton accepting thecomplement of L.
Deterministic case:The cost is 4n, for any input alphabet.[Geffert, Mereghetti, Pighizzini 2007](note that 2dfa’s can have infinite computations)Nondeterministic case:The cost is polynomial for a unary alphabet.[Geffert, Mereghetti, Pighizzini 2007]
What about the complementation of 2nfa over nonunaryalphabets?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 151: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/151.jpg)
Complementation of two-way automata
ProblemGiven a two-way automaton with n states accepting a languageL, find the cost in term of states, of an automaton accepting thecomplement of L.
Deterministic case:The cost is 4n, for any input alphabet.[Geffert, Mereghetti, Pighizzini 2007](note that 2dfa’s can have infinite computations)Nondeterministic case:The cost is polynomial for a unary alphabet.[Geffert, Mereghetti, Pighizzini 2007]
What about the complementation of 2nfa over nonunaryalphabets?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 152: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/152.jpg)
Complementation of two-way automata
ProblemGiven a two-way automaton with n states accepting a languageL, find the cost in term of states, of an automaton accepting thecomplement of L.
Deterministic case:The cost is 4n, for any input alphabet.[Geffert, Mereghetti, Pighizzini 2007](note that 2dfa’s can have infinite computations)Nondeterministic case:The cost is polynomial for a unary alphabet.[Geffert, Mereghetti, Pighizzini 2007]
What about the complementation of 2nfa over nonunaryalphabets?
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 153: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/153.jpg)
Complementation of two-way automata
Relationship with the Sakoda&Sipser question
f (n) := the cost of the simulation of a n-state 2nfa by a 2dfa.
Given an n-state 2nfa accepting L we can find:a 2dfa accepting L with f (n) statesa 2nfa (actually a 2dfa) accepting Lc with 4f (n) states
Hence:the complementation of 2nfa’s costs no more than theirdeterminization.
TheoremIf the complementation of 2nfa’s requires an exponentialnumber of states then the gap between 2nfa’s and 2dfa’s isexponential.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 154: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/154.jpg)
Complementation of two-way automata
Relationship with the Sakoda&Sipser question
f (n) := the cost of the simulation of a n-state 2nfa by a 2dfa.
Given an n-state 2nfa accepting L we can find:a 2dfa accepting L with f (n) statesa 2nfa (actually a 2dfa) accepting Lc with 4f (n) states
Hence:the complementation of 2nfa’s costs no more than theirdeterminization.
TheoremIf the complementation of 2nfa’s requires an exponentialnumber of states then the gap between 2nfa’s and 2dfa’s isexponential.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 155: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/155.jpg)
Complementation of two-way automata
Relationship with the Sakoda&Sipser question
f (n) := the cost of the simulation of a n-state 2nfa by a 2dfa.
Given an n-state 2nfa accepting L we can find:a 2dfa accepting L with f (n) statesa 2nfa (actually a 2dfa) accepting Lc with 4f (n) states
Hence:the complementation of 2nfa’s costs no more than theirdeterminization.
TheoremIf the complementation of 2nfa’s requires an exponentialnumber of states then the gap between 2nfa’s and 2dfa’s isexponential.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 156: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/156.jpg)
Complementation of two-way automata
Relationship with the Sakoda&Sipser question
f (n) := the cost of the simulation of a n-state 2nfa by a 2dfa.
Given an n-state 2nfa accepting L we can find:a 2dfa accepting L with f (n) statesa 2nfa (actually a 2dfa) accepting Lc with 4f (n) states
Hence:the complementation of 2nfa’s costs no more than theirdeterminization.
TheoremIf the complementation of 2nfa’s requires an exponentialnumber of states then the gap between 2nfa’s and 2dfa’s isexponential.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 157: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/157.jpg)
Complementation of two-way automata
Relationship with the Sakoda&Sipser question
f (n) := the cost of the simulation of a n-state 2nfa by a 2dfa.
Given an n-state 2nfa accepting L we can find:a 2dfa accepting L with f (n) statesa 2nfa (actually a 2dfa) accepting Lc with 4f (n) states
Hence:the complementation of 2nfa’s costs no more than theirdeterminization.
TheoremIf the complementation of 2nfa’s requires an exponentialnumber of states then the gap between 2nfa’s and 2dfa’s isexponential.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 158: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/158.jpg)
Complementation of two-way automata
Relationship with the Sakoda&Sipser question
f (n) := the cost of the simulation of a n-state 2nfa by a 2dfa.
Given an n-state 2nfa accepting L we can find:a 2dfa accepting L with f (n) statesa 2nfa (actually a 2dfa) accepting Lc with 4f (n) states
Hence:the complementation of 2nfa’s costs no more than theirdeterminization.
TheoremIf the complementation of 2nfa’s requires an exponentialnumber of states then the gap between 2nfa’s and 2dfa’s isexponential.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 159: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/159.jpg)
Complementation of one-way automata
Deterministic case: trivial
Nondeterministic case:Worst case: [Jiraskova, 2005]For each integer n ≥ 1 there exists a language L acceptedby a 1nfa with n states such that each 1nfa accepting Lc
needs 2n states.
Best case: [Mera, Pighizzini, 2005]For each integer n ≥ 1 there exists a language L such that:– the smallest 1nfa accepting L has n states– the minimum 1dfa’s accepting L and Lc have 2n states– the smallest 1nfa accepting Lc has at most n + 1 states.
Remarks:L is a witness of the gap between 1nfa’s and 1dfa’s.Both L and Lc have “small” 1nfa but “large” 1dfa’s.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 160: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/160.jpg)
Complementation of one-way automata
Deterministic case: trivial
Nondeterministic case:Worst case: [Jiraskova, 2005]For each integer n ≥ 1 there exists a language L acceptedby a 1nfa with n states such that each 1nfa accepting Lc
needs 2n states.
Best case: [Mera, Pighizzini, 2005]For each integer n ≥ 1 there exists a language L such that:– the smallest 1nfa accepting L has n states– the minimum 1dfa’s accepting L and Lc have 2n states– the smallest 1nfa accepting Lc has at most n + 1 states.
Remarks:L is a witness of the gap between 1nfa’s and 1dfa’s.Both L and Lc have “small” 1nfa but “large” 1dfa’s.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 161: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/161.jpg)
Complementation of one-way automata
Deterministic case: trivial
Nondeterministic case:Worst case: [Jiraskova, 2005]For each integer n ≥ 1 there exists a language L acceptedby a 1nfa with n states such that each 1nfa accepting Lc
needs 2n states.
Best case: [Mera, Pighizzini, 2005]For each integer n ≥ 1 there exists a language L such that:– the smallest 1nfa accepting L has n states– the minimum 1dfa’s accepting L and Lc have 2n states– the smallest 1nfa accepting Lc has at most n + 1 states.
Remarks:L is a witness of the gap between 1nfa’s and 1dfa’s.Both L and Lc have “small” 1nfa but “large” 1dfa’s.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 162: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/162.jpg)
Complementation of one-way automata
Deterministic case: trivial
Nondeterministic case:Worst case: [Jiraskova, 2005]For each integer n ≥ 1 there exists a language L acceptedby a 1nfa with n states such that each 1nfa accepting Lc
needs 2n states.
Best case: [Mera, Pighizzini, 2005]For each integer n ≥ 1 there exists a language L such that:– the smallest 1nfa accepting L has n states– the minimum 1dfa’s accepting L and Lc have 2n states– the smallest 1nfa accepting Lc has at most n + 1 states.
Remarks:L is a witness of the gap between 1nfa’s and 1dfa’s.Both L and Lc have “small” 1nfa but “large” 1dfa’s.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 163: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/163.jpg)
Complementation of one-way automata
Deterministic case: trivial
Nondeterministic case:Worst case: [Jiraskova, 2005]For each integer n ≥ 1 there exists a language L acceptedby a 1nfa with n states such that each 1nfa accepting Lc
needs 2n states.
Best case: [Mera, Pighizzini, 2005]For each integer n ≥ 1 there exists a language L such that:– the smallest 1nfa accepting L has n states– the minimum 1dfa’s accepting L and Lc have 2n states– the smallest 1nfa accepting Lc has at most n + 1 states.
Remarks:L is a witness of the gap between 1nfa’s and 1dfa’s.Both L and Lc have “small” 1nfa but “large” 1dfa’s.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 164: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/164.jpg)
Complementation of one-way automata
Deterministic case: trivial
Nondeterministic case:Worst case: [Jiraskova, 2005]For each integer n ≥ 1 there exists a language L acceptedby a 1nfa with n states such that each 1nfa accepting Lc
needs 2n states.
Best case: [Mera, Pighizzini, 2005]For each integer n ≥ 1 there exists a language L such that:– the smallest 1nfa accepting L has n states– the minimum 1dfa’s accepting L and Lc have 2n states– the smallest 1nfa accepting Lc has at most n + 1 states.
Remarks:L is a witness of the gap between 1nfa’s and 1dfa’s.Both L and Lc have “small” 1nfa but “large” 1dfa’s.
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 165: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/165.jpg)
Complementation of 1nfa: unary case
The last result does not hold in the case ofunary languages:
For each unary language L such that:L is accepted by a 1nfa with n states
L the minimum 1dfa accepting L has eO(√
n log n) statesThen: [Mera, Pighizzini, 2005]
Each 1nfa accepting Lc should have at least n states.
Hence, if L has a “small” 1nfa and a “large” 1dfa, i.e., it is awitness of the gap between unary 1nfa’s and 1dfa’s, then thenondeterminism does not help in the recognition of Lc .
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 166: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/166.jpg)
Complementation of 1nfa: unary case
The last result does not hold in the case ofunary languages:
For each unary language L such that:L is accepted by a 1nfa with n states
L the minimum 1dfa accepting L has eO(√
n log n) statesThen: [Mera, Pighizzini, 2005]
Each 1nfa accepting Lc should have at least n states.
Hence, if L has a “small” 1nfa and a “large” 1dfa, i.e., it is awitness of the gap between unary 1nfa’s and 1dfa’s, then thenondeterminism does not help in the recognition of Lc .
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 167: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/167.jpg)
Complementation of 1nfa: unary case
The last result does not hold in the case ofunary languages:
For each unary language L such that:L is accepted by a 1nfa with n states
L the minimum 1dfa accepting L has eO(√
n log n) statesThen: [Mera, Pighizzini, 2005]
Each 1nfa accepting Lc should have at least n states.
Hence, if L has a “small” 1nfa and a “large” 1dfa, i.e., it is awitness of the gap between unary 1nfa’s and 1dfa’s, then thenondeterminism does not help in the recognition of Lc .
Giovanni Pighizzini Descriptional complexity of automata and languages
![Page 168: Descriptional complexity of automata and languageshomes.di.unimi.it/~pighizzi/talks/2008/logica.pdfDescriptional complexity of automata and languages Giovanni Pighizzini Dipartimento](https://reader034.vdocument.in/reader034/viewer/2022042108/5e879eb8580c447f76247791/html5/thumbnails/168.jpg)
Complementation of 1nfa: unary case
The last result does not hold in the case ofunary languages:
For each unary language L such that:L is accepted by a 1nfa with n states
L the minimum 1dfa accepting L has eO(√
n log n) statesThen: [Mera, Pighizzini, 2005]
Each 1nfa accepting Lc should have at least n states.
Hence, if L has a “small” 1nfa and a “large” 1dfa, i.e., it is awitness of the gap between unary 1nfa’s and 1dfa’s, then thenondeterminism does not help in the recognition of Lc .
Giovanni Pighizzini Descriptional complexity of automata and languages