boolean circuits: ein alternatives rechenmodell€¦ · 1 simon bischof, philipp christian...
TRANSCRIPT
![Page 1: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/1.jpg)
1 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Institut fur Theoretische Informatik
Boolean circuits: ein alternatives RechenmodellSimon Bischof, Philipp Christian Loewner
KIT – Universitat des Landes Baden-Wurttemberg undnationales Forschungszentrum in der Helmholtz-Gemeinschaft www.kit.edu
![Page 2: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/2.jpg)
2 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Motivation
![Page 3: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/3.jpg)
Mein erster Boolean Circuit
3 Simon Bischof, Philipp Christian Loewner:Boolean circuits
𝑥𝟏 𝑥𝟑 𝑥𝟐
¬
∧ ∧
∨
![Page 4: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/4.jpg)
Mein erster Boolean Circuit
3 Simon Bischof, Philipp Christian Loewner:Boolean circuits
𝑥𝟏 𝑥𝟑 𝑥𝟐
¬
∧ ∧
∨
C(x1, x2, x3) = x1x2 ∨ x2x3
![Page 5: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/5.jpg)
Mein erster Boolean Circuit
3 Simon Bischof, Philipp Christian Loewner:Boolean circuits
𝑥𝟏 𝑥𝟑 𝑥𝟐
¬
∧ ∧
∨
Eingabeknoten
Ausgabeknoten
C(x1, x2, x3) = x1x2 ∨ x2x3
![Page 6: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/6.jpg)
Mein erster Boolean Circuit
3 Simon Bischof, Philipp Christian Loewner:Boolean circuits
𝑥𝟏 𝑥𝟑 𝑥𝟐
¬
∧ ∧
∨
1 0
1
0
1 0
1
Auswertung fur x := (x1, x2, x3) = (0,0,1) ergibt C(x) = 1
![Page 7: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/7.jpg)
Eigenschaften
4 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits entsprechen Schaltnetzen, aber nicht Schaltwerken -mussen azyklisch sein
Boolean Circuits sind Boolesche Formeln, wenn jeder Knoten hochstensAusgangsgrad 1 hat.
![Page 8: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/8.jpg)
Benutzung fur Sprachen
5 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Idee: Benutzung zur Erkennung von Wortern
Circuit entscheidet Worter seiner Eingabelange⇒ ein Circuit pro EingabegroßeFolge von Circuits pro Sprache
Folge {Cn}n∈N von Circuits heißt Circuit-Familie.
Dabei hat jeder Circuit Cn die Eingabegroße n
![Page 9: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/9.jpg)
Benutzung fur Sprachen
5 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Idee: Benutzung zur Erkennung von Wortern
Circuit entscheidet Worter seiner Eingabelange⇒ ein Circuit pro EingabegroßeFolge von Circuits pro Sprache
Folge {Cn}n∈N von Circuits heißt Circuit-Familie.
Dabei hat jeder Circuit Cn die Eingabegroße n
![Page 10: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/10.jpg)
Beispiel Circuit-Familie
6 Simon Bischof, Philipp Christian Loewner:Boolean circuits
𝑥𝟏 𝑥𝟑 𝑥𝟐
¬
∧ ∧
∨
𝑥𝟏
𝑥𝟐 𝑥𝟏
¬
∧
∧
. . .
![Page 11: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/11.jpg)
Beispiel Circuit-Familie
7 Simon Bischof, Philipp Christian Loewner:Boolean circuits
L := {1n | n ∈N}
![Page 12: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/12.jpg)
Interessante Eigenschaften
8 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits konnen alle Sprachen entscheiden die TMsentscheiden konnenTuringmaschinen konnen Boolean Circuits simulieren
Boolean Circuits sind aber machtiger als TMsbetrachte unare Spracheneindeutiges Wort wn fur jede Wortlange⇒ wn ∈ L kann hart in Cn eincodiert werden⇒ Circuit-Familien konnen unentscheidbare Sprachen entscheidenBeispiel: {1n : Mn akzeptiert 1n}
![Page 13: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/13.jpg)
Interessante Eigenschaften
8 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits konnen alle Sprachen entscheiden die TMsentscheiden konnenTuringmaschinen konnen Boolean Circuits simulierenBoolean Circuits sind aber machtiger als TMs
betrachte unare Spracheneindeutiges Wort wn fur jede Wortlange⇒ wn ∈ L kann hart in Cn eincodiert werden⇒ Circuit-Familien konnen unentscheidbare Sprachen entscheidenBeispiel: {1n : Mn akzeptiert 1n}
![Page 14: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/14.jpg)
Interessante Eigenschaften
8 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits konnen alle Sprachen entscheiden die TMsentscheiden konnenTuringmaschinen konnen Boolean Circuits simulierenBoolean Circuits sind aber machtiger als TMs
betrachte unare Spracheneindeutiges Wort wn fur jede Wortlange
⇒ wn ∈ L kann hart in Cn eincodiert werden⇒ Circuit-Familien konnen unentscheidbare Sprachen entscheidenBeispiel: {1n : Mn akzeptiert 1n}
![Page 15: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/15.jpg)
Interessante Eigenschaften
8 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits konnen alle Sprachen entscheiden die TMsentscheiden konnenTuringmaschinen konnen Boolean Circuits simulierenBoolean Circuits sind aber machtiger als TMs
betrachte unare Spracheneindeutiges Wort wn fur jede Wortlange⇒ wn ∈ L kann hart in Cn eincodiert werden
⇒ Circuit-Familien konnen unentscheidbare Sprachen entscheidenBeispiel: {1n : Mn akzeptiert 1n}
![Page 16: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/16.jpg)
Interessante Eigenschaften
8 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits konnen alle Sprachen entscheiden die TMsentscheiden konnenTuringmaschinen konnen Boolean Circuits simulierenBoolean Circuits sind aber machtiger als TMs
betrachte unare Spracheneindeutiges Wort wn fur jede Wortlange⇒ wn ∈ L kann hart in Cn eincodiert werden⇒ Circuit-Familien konnen unentscheidbare Sprachen entscheidenBeispiel: {1n : Mn akzeptiert 1n}
![Page 17: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/17.jpg)
Weitere Vorgehensweise
9 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition von KomplexitatsklassenEinfuhrung sinnvoller BeschrankungenWozu kann man Circuits denn nun verwenden?
![Page 18: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/18.jpg)
Weitere Vorgehensweise
10 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition von KomplexitatsklassenEinfuhrung sinnvoller BeschrankungenWozu kann man Circuits denn nun verwenden?
![Page 19: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/19.jpg)
11 Simon Bischof, Philipp Christian Loewner:Boolean circuits
P/poly und P
![Page 20: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/20.jpg)
Boolean circuits - Große
12 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Große eines einzelnen Circuits Cn: Skalar|Cn| := Anzahl der Knoten in Cn
Große einer Familie {Cn}n∈N: FunktionT (n), wobei |Cn| ≤ T (n) (n ∈N)
Klassifizierung einer Sprache L ⊆ {0,1}+L ∈ SIZE(T (n)), falls eine Circuit-Familie der Große T (n) existiert diegenau L entscheidet
![Page 21: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/21.jpg)
Anwendung auf vorige Beispiele
13 Simon Bischof, Philipp Christian Loewner:Boolean circuits
L := {1n | n ∈N}:
Folgerung: L ∈ SIZE(2n)
![Page 22: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/22.jpg)
Anwendung auf vorige Beispiele
13 Simon Bischof, Philipp Christian Loewner:Boolean circuits
L := {1n | n ∈N}:
Folgerung: L ∈ SIZE(2n)
![Page 23: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/23.jpg)
Anwendung auf vorige Beispiele
14 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Sei n ∈N, f : {0,1}n → {0,1} beliebig.
f lasst sich in konjunktiver Normalform darstellenCircuit dafur hat hochstens n · 2n KnotenFolgerung: ∀L ⊆ {0,1}+ : L ∈ SIZE(n · 2n)
![Page 24: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/24.jpg)
Anwendung auf vorige Beispiele
14 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Sei n ∈N, f : {0,1}n → {0,1} beliebig.f lasst sich in konjunktiver Normalform darstellen
Circuit dafur hat hochstens n · 2n KnotenFolgerung: ∀L ⊆ {0,1}+ : L ∈ SIZE(n · 2n)
![Page 25: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/25.jpg)
Anwendung auf vorige Beispiele
14 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Sei n ∈N, f : {0,1}n → {0,1} beliebig.f lasst sich in konjunktiver Normalform darstellenCircuit dafur hat hochstens n · 2n Knoten
Folgerung: ∀L ⊆ {0,1}+ : L ∈ SIZE(n · 2n)
![Page 26: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/26.jpg)
Anwendung auf vorige Beispiele
14 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Sei n ∈N, f : {0,1}n → {0,1} beliebig.f lasst sich in konjunktiver Normalform darstellenCircuit dafur hat hochstens n · 2n KnotenFolgerung: ∀L ⊆ {0,1}+ : L ∈ SIZE(n · 2n)
![Page 27: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/27.jpg)
Die Klasse P/poly
15 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bei Turingmaschinen haben wir mit P effizient losbare Sprachenbetrachtet.
Wir untersuchen nun, welche Komplexitatsklassen Boolean Circuits beider Einschrankung auf “kleine“ Circuits liefern.
P/poly :=⋃
c SIZE(nc)
Klasse der durch polynomiell große Familien von Circuitsentscheidbaren SprachenSatz: es gilt P ( P/poly .
![Page 28: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/28.jpg)
Die Klasse P/poly
15 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bei Turingmaschinen haben wir mit P effizient losbare Sprachenbetrachtet.Wir untersuchen nun, welche Komplexitatsklassen Boolean Circuits beider Einschrankung auf “kleine“ Circuits liefern.
P/poly :=⋃
c SIZE(nc)
Klasse der durch polynomiell große Familien von Circuitsentscheidbaren SprachenSatz: es gilt P ( P/poly .
![Page 29: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/29.jpg)
Die Klasse P/poly
15 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bei Turingmaschinen haben wir mit P effizient losbare Sprachenbetrachtet.Wir untersuchen nun, welche Komplexitatsklassen Boolean Circuits beider Einschrankung auf “kleine“ Circuits liefern.
P/poly :=⋃
c SIZE(nc)
Klasse der durch polynomiell große Familien von Circuitsentscheidbaren SprachenSatz: es gilt P ( P/poly .
![Page 30: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/30.jpg)
Die Klasse P/poly
15 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bei Turingmaschinen haben wir mit P effizient losbare Sprachenbetrachtet.Wir untersuchen nun, welche Komplexitatsklassen Boolean Circuits beider Einschrankung auf “kleine“ Circuits liefern.
P/poly :=⋃
c SIZE(nc)
Klasse der durch polynomiell große Familien von Circuitsentscheidbaren Sprachen
Satz: es gilt P ( P/poly .
![Page 31: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/31.jpg)
Die Klasse P/poly
15 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bei Turingmaschinen haben wir mit P effizient losbare Sprachenbetrachtet.Wir untersuchen nun, welche Komplexitatsklassen Boolean Circuits beider Einschrankung auf “kleine“ Circuits liefern.
P/poly :=⋃
c SIZE(nc)
Klasse der durch polynomiell große Familien von Circuitsentscheidbaren SprachenSatz: es gilt P ( P/poly .
![Page 32: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/32.jpg)
Begrundung fur P ( P/poly
16 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 1: P 6= P/poly
Erinnerung: L := {1n : Mn akzeptiert 1n} unentscheidbarjede unare Sprache hat Circuit-Familie linearer Große⇒ L ∈ SIZE(O(n))offensichtlich aber L /∈ P
Hinweis: Problem der Unentscheidbarkeit ist eigentlich nur verlagert indie Generierung der Circuit-Familie.Im nachsten Kapitel: Beschrankung auf eine Klasse die gleich P ist.
![Page 33: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/33.jpg)
Begrundung fur P ( P/poly
17 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
Jede Sprache L ∈ P lasst sich in polynomieller Zeit von einer TM Mentscheiden.
Wie konnen wir das auf Circuits ubertragen?
Probleme:
TM andert BandinhaltTM trifft Entscheidungen basierend auf BandinhaltCircuits konnen ihre Eingabe nicht verandernBetrachtung der Große von Circuits, aber Laufzeit von TMs
Losung: Betrachtung von oblivious TMs
![Page 34: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/34.jpg)
Begrundung fur P ( P/poly
17 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
Jede Sprache L ∈ P lasst sich in polynomieller Zeit von einer TM Mentscheiden.
Wie konnen wir das auf Circuits ubertragen?
Probleme:
TM andert BandinhaltTM trifft Entscheidungen basierend auf BandinhaltCircuits konnen ihre Eingabe nicht verandernBetrachtung der Große von Circuits, aber Laufzeit von TMs
Losung: Betrachtung von oblivious TMs
![Page 35: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/35.jpg)
Begrundung fur P ( P/poly
18 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
L ∈ P⇒ ex. TM, die L in polynomieller Zeit entscheidet⇒ ex. oblivious TM, die L in polynomieller Zeit entscheidet⇒ ex. Circuit-Familie polynomieller Große, die L entscheidet
![Page 36: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/36.jpg)
Exkurs: Oblivious TM
19 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Normale TuringmaschinePosition auf Band abhangig von Eingabe und Schrittnummer
0 1 1 1 0 1 1 1 1 0 0 1 1 1 0 0 1 1 1 0
0 0 1 1 0 1 0 1 1 0 1 0 1 1 0 1 0 0 1 0
![Page 37: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/37.jpg)
Exkurs: Oblivious TM
19 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Oblivious TuringmaschinePosition auf Band nur abhangig von Eingabelange und Schrittnummer
0 1 1 1 0 1 1 1 1 0 1 1 1 1 0 1 1 1 1 0
0 0 1 1 0 1 0 1 1 0 1 0 1 1 0 1 0 0 1 0
![Page 38: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/38.jpg)
Begrundung fur P ( P/poly
20 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
L ∈ P⇒ ex. TM, die L in polynomieller Zeit entscheidet⇒ ex. oblivious TM, die L in polynomieller Zeit entscheidet⇒ ex. Circuit-Familie polynomieller Große, die L entscheidet
TM M1 existiert mit pol. Laufzeit T (n)erstelle daraus eine oblivious TM M2
⇒ Laufzeit in O(T 2(n))⇒ also insgesamt polynomielle Laufzeit
![Page 39: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/39.jpg)
Begrundung fur P ( P/poly
20 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
L ∈ P⇒ ex. TM, die L in polynomieller Zeit entscheidet⇒ ex. oblivious TM, die L in polynomieller Zeit entscheidet⇒ ex. Circuit-Familie polynomieller Große, die L entscheidet
![Page 40: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/40.jpg)
Begrundung fur P ( P/poly
20 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
L ∈ P⇒ ex. TM, die L in polynomieller Zeit entscheidet⇒ ex. oblivious TM, die L in polynomieller Zeit entscheidet⇒ ex. Circuit-Familie polynomieller Große, die L entscheidet
oblivious TM M2 existiert mit pol. Laufzeitfeste Laufzeit pro Eingabegroßeakzeptiert falls im letzten Schritt erreichter Zustand akzeptierend istIdee: Codiere Bandalphabet und Zustandsmenge in Knoten
![Page 41: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/41.jpg)
Begrundung fur P ( P/poly
20 Simon Bischof, Philipp Christian Loewner:Boolean circuits
𝑧𝑖 =𝑍𝑢𝑠𝑡𝑎𝑛𝑑 𝑛𝑎𝑐ℎ 𝑆𝑐ℎ𝑟𝑖𝑡𝑡 𝑖𝑔𝑒𝑠𝑐ℎ𝑟𝑖𝑒𝑏𝑒𝑛𝑒𝑠 𝑍𝑒𝑖𝑐ℎ𝑒𝑛
z.B. 𝑧1 =𝑞2
1
0 1 1 1 0 1 1 1 1 0 1 1 1 1 0 1 1 1 1 0
0 0 1 1 0 1 0 1 1 0 1 0 1 1 0 1 0 0 1 0
![Page 42: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/42.jpg)
Begrundung fur P ( P/poly
20 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Teil 2: P ⊆ P/poly
L ∈ P⇒ ex. TM, die L in polynomieller Zeit entscheidet⇒ ex. oblivious TM, die L in polynomieller Zeit entscheidet⇒ ex. Circuit-Familie polynomieller Große, die L entscheidet
Idee: Codiere Bandalphabet und Zustandsmenge in Knotenkonstantes Alphabet, konstante Zustandsmenge⇒ Anzahl der Knoten fur beliebige Operationen ist konstantbeschrankt⇒ jeder Berechnungsschritt lasst sich durch einen Teilcircuit konstantbeschrankter Große darstellen⇒ fur polynomielle Anzahl Schritte: Polynomielle Anzahl von Knoten
![Page 43: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/43.jpg)
Begrundung fur P ( P/poly
21 Simon Bischof, Philipp Christian Loewner:Boolean circuits
geschriebenes Zeichen neuer Zustand
Zustand aus 𝑧𝑖−1
Zeichen aus 𝑧𝑗
Schritt i
![Page 44: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/44.jpg)
Weitere Vorgehensweise
22 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition von KomplexitatsklassenEinfuhrung sinnvoller BeschrankungenWozu kann man Circuits denn nun verwenden?
![Page 45: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/45.jpg)
23 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Uniformitat
![Page 46: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/46.jpg)
Motivation zur Uniformitat
24 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Wie konnen wirboolean circuits einschranken um sie auf die Machtigkeit von TMs zubringen?TMs erweitern um sie auf die Machtigkeit von boolean circuits zubringen?
![Page 47: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/47.jpg)
P-uniform, logspace-uniform
25 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Eine Familie {Cn} von Circuits heißt P-uniform, wenn es einepolynomielle Turingmaschine gibt, die aus der Eingabe 1n dieBeschreibung von Cn berechnet.
Durch die Einschrankung auf P-uniforme Circuits wird P/poly zu P .
Eine Circuit-Familie heißt logspace-uniform, wenn die FunktionenSIZE(n) = |Cn|TYPE(n, i) (Typ des i-ten Knoten von Cn)EDGE(n, i , j) (1 falls es eine Kante vom i-ten zum j-ten Knoten von Cn gibt)
im Sinne der gangigen Definition von logspace berechenbar sind.Logspace-Uniformitat ist zusatzlich zu P-Uniformitat keine echteEinschrankung.Eine Sprache L besitzt polynomiell große, logspace-uniforme Circuits⇔ L ∈ P .
![Page 48: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/48.jpg)
P-uniform, logspace-uniform
25 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Eine Familie {Cn} von Circuits heißt P-uniform, wenn es einepolynomielle Turingmaschine gibt, die aus der Eingabe 1n dieBeschreibung von Cn berechnet.Durch die Einschrankung auf P-uniforme Circuits wird P/poly zu P .
Eine Circuit-Familie heißt logspace-uniform, wenn die FunktionenSIZE(n) = |Cn|TYPE(n, i) (Typ des i-ten Knoten von Cn)EDGE(n, i , j) (1 falls es eine Kante vom i-ten zum j-ten Knoten von Cn gibt)
im Sinne der gangigen Definition von logspace berechenbar sind.Logspace-Uniformitat ist zusatzlich zu P-Uniformitat keine echteEinschrankung.Eine Sprache L besitzt polynomiell große, logspace-uniforme Circuits⇔ L ∈ P .
![Page 49: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/49.jpg)
P-uniform, logspace-uniform
25 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Eine Familie {Cn} von Circuits heißt P-uniform, wenn es einepolynomielle Turingmaschine gibt, die aus der Eingabe 1n dieBeschreibung von Cn berechnet.Durch die Einschrankung auf P-uniforme Circuits wird P/poly zu P .
Eine Circuit-Familie heißt logspace-uniform, wenn die Funktionen
SIZE(n) = |Cn|TYPE(n, i) (Typ des i-ten Knoten von Cn)EDGE(n, i , j) (1 falls es eine Kante vom i-ten zum j-ten Knoten von Cn gibt)
im Sinne der gangigen Definition von logspace berechenbar sind.Logspace-Uniformitat ist zusatzlich zu P-Uniformitat keine echteEinschrankung.Eine Sprache L besitzt polynomiell große, logspace-uniforme Circuits⇔ L ∈ P .
![Page 50: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/50.jpg)
P-uniform, logspace-uniform
25 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Eine Familie {Cn} von Circuits heißt P-uniform, wenn es einepolynomielle Turingmaschine gibt, die aus der Eingabe 1n dieBeschreibung von Cn berechnet.Durch die Einschrankung auf P-uniforme Circuits wird P/poly zu P .
Eine Circuit-Familie heißt logspace-uniform, wenn die FunktionenSIZE(n) = |Cn|TYPE(n, i) (Typ des i-ten Knoten von Cn)EDGE(n, i , j) (1 falls es eine Kante vom i-ten zum j-ten Knoten von Cn gibt)
im Sinne der gangigen Definition von logspace berechenbar sind.Logspace-Uniformitat ist zusatzlich zu P-Uniformitat keine echteEinschrankung.Eine Sprache L besitzt polynomiell große, logspace-uniforme Circuits⇔ L ∈ P .
![Page 51: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/51.jpg)
P-uniform, logspace-uniform
25 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Eine Familie {Cn} von Circuits heißt P-uniform, wenn es einepolynomielle Turingmaschine gibt, die aus der Eingabe 1n dieBeschreibung von Cn berechnet.Durch die Einschrankung auf P-uniforme Circuits wird P/poly zu P .
Eine Circuit-Familie heißt logspace-uniform, wenn die FunktionenSIZE(n) = |Cn|TYPE(n, i) (Typ des i-ten Knoten von Cn)EDGE(n, i , j) (1 falls es eine Kante vom i-ten zum j-ten Knoten von Cn gibt)
im Sinne der gangigen Definition von logspace berechenbar sind.
Logspace-Uniformitat ist zusatzlich zu P-Uniformitat keine echteEinschrankung.Eine Sprache L besitzt polynomiell große, logspace-uniforme Circuits⇔ L ∈ P .
![Page 52: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/52.jpg)
P-uniform, logspace-uniform
25 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Eine Familie {Cn} von Circuits heißt P-uniform, wenn es einepolynomielle Turingmaschine gibt, die aus der Eingabe 1n dieBeschreibung von Cn berechnet.Durch die Einschrankung auf P-uniforme Circuits wird P/poly zu P .
Eine Circuit-Familie heißt logspace-uniform, wenn die FunktionenSIZE(n) = |Cn|TYPE(n, i) (Typ des i-ten Knoten von Cn)EDGE(n, i , j) (1 falls es eine Kante vom i-ten zum j-ten Knoten von Cn gibt)
im Sinne der gangigen Definition von logspace berechenbar sind.Logspace-Uniformitat ist zusatzlich zu P-Uniformitat keine echteEinschrankung.Eine Sprache L besitzt polynomiell große, logspace-uniforme Circuits⇔ L ∈ P .
![Page 53: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/53.jpg)
Turingmaschinen mit advice strings
26 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Normale TM
0 1 1 0 … 0 1
n
![Page 54: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/54.jpg)
Turingmaschinen mit advice strings
26 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Advice TM fragt Blackbox
0 1 1 0 … 0 1
n
?
![Page 55: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/55.jpg)
Turingmaschinen mit advice strings
26 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Advice TM erhalt advice von Blackbox
0 1 1 0 … 0 1
n
? 1 0 1 0 0 … 0 1 0
a(n)
![Page 56: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/56.jpg)
Turingmaschinen mit advice strings
26 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bemerkungen:1. jede unare Sprache lasst sich mit einem bit advice in linearer Zeit
entscheiden2. P/poly : Sprachen, die mit polynomiellem advice in polynomieller Zeit
entscheidbar sind
![Page 57: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/57.jpg)
Die Sprache CKT-SAT
27 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Ein Wort c ist in CKT-SAT, wenn der durch c beschriebene Circuit Ceine erfullende Belegung besitzt.
CKT-SAT∈ NP . (Warum?)CKT-SAT ist auch NP-schwer, also NP-vollstandig.CKT-SAT ∝p 3SAT.
![Page 58: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/58.jpg)
Die Sprache CKT-SAT
27 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Ein Wort c ist in CKT-SAT, wenn der durch c beschriebene Circuit Ceine erfullende Belegung besitzt.CKT-SAT∈ NP .
(Warum?)CKT-SAT ist auch NP-schwer, also NP-vollstandig.CKT-SAT ∝p 3SAT.
![Page 59: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/59.jpg)
Die Sprache CKT-SAT
27 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Ein Wort c ist in CKT-SAT, wenn der durch c beschriebene Circuit Ceine erfullende Belegung besitzt.CKT-SAT∈ NP . (Warum?)
CKT-SAT ist auch NP-schwer, also NP-vollstandig.CKT-SAT ∝p 3SAT.
![Page 60: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/60.jpg)
Die Sprache CKT-SAT
27 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Ein Wort c ist in CKT-SAT, wenn der durch c beschriebene Circuit Ceine erfullende Belegung besitzt.CKT-SAT∈ NP . (Warum?)CKT-SAT ist auch NP-schwer, also NP-vollstandig.
CKT-SAT ∝p 3SAT.
![Page 61: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/61.jpg)
Die Sprache CKT-SAT
27 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Ein Wort c ist in CKT-SAT, wenn der durch c beschriebene Circuit Ceine erfullende Belegung besitzt.CKT-SAT∈ NP . (Warum?)CKT-SAT ist auch NP-schwer, also NP-vollstandig.CKT-SAT ∝p 3SAT.
![Page 62: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/62.jpg)
Die Sprache CKT-SAT
28 Simon Bischof, Philipp Christian Loewner:Boolean circuits
? Eingabe
TM M
ja / nein
x u
![Page 63: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/63.jpg)
Die Sprache CKT-SAT
28 Simon Bischof, Philipp Christian Loewner:Boolean circuits
? Eingabe
TM M
ja / nein
x
Circuit
u
![Page 64: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/64.jpg)
Die Sprache CKT-SAT
28 Simon Bischof, Philipp Christian Loewner:Boolean circuits
? Eingabe
TM M
ja / nein
x u
Circuit
Eingabe
u
ja / nein
![Page 65: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/65.jpg)
Weitere Vorgehensweise
29 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition von KomplexitatsklassenEinfuhrung sinnvoller BeschrankungenWozu kann man Circuits denn nun verwenden?
Neue Ansatze zur Untersuchung von KomplexitatsklassenBessere Beschreibung paralleler Algorithmen
![Page 66: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/66.jpg)
Weitere Vorgehensweise
29 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition von KomplexitatsklassenEinfuhrung sinnvoller BeschrankungenWozu kann man Circuits denn nun verwenden?
Neue Ansatze zur Untersuchung von KomplexitatsklassenBessere Beschreibung paralleler Algorithmen
![Page 67: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/67.jpg)
30 Simon Bischof, Philipp Christian Loewner:Boolean circuits
P/poly und NP
![Page 68: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/68.jpg)
Zur Erinnerung: Polynomielle Hierarchie
31 Simon Bischof, Philipp Christian Loewner:Boolean circuits
(Bild aus Wikipedia)
![Page 69: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/69.jpg)
Satz von Meyer
32 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls EXPTIME ⊆ P/poly , dann gilt EXPTIME = Σp2.
Folgerung: Falls EXPTIME ⊆ P/poly , dann gilt P 6= NP . Beweis:
1. aus EXPTIME ⊆ P/poly folgt EXPTIME = Σp2
2. ware P = NP , so ware P = Σp2
3. ⇒ P = EXPTIME4. dies widerspricht dem Zeithierarchie-Theorem
![Page 70: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/70.jpg)
Satz von Meyer
32 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls EXPTIME ⊆ P/poly , dann gilt EXPTIME = Σp2.
Folgerung: Falls EXPTIME ⊆ P/poly , dann gilt P 6= NP . Beweis:
1. aus EXPTIME ⊆ P/poly folgt EXPTIME = Σp2
2. ware P = NP , so ware P = Σp2
3. ⇒ P = EXPTIME4. dies widerspricht dem Zeithierarchie-Theorem
![Page 71: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/71.jpg)
Satz von Meyer
32 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls EXPTIME ⊆ P/poly , dann gilt EXPTIME = Σp2.
Folgerung: Falls EXPTIME ⊆ P/poly , dann gilt P 6= NP . Beweis:
1. aus EXPTIME ⊆ P/poly folgt EXPTIME = Σp2
2. ware P = NP , so ware P = Σp2
3. ⇒ P = EXPTIME4. dies widerspricht dem Zeithierarchie-Theorem
![Page 72: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/72.jpg)
Satz von Meyer
32 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls EXPTIME ⊆ P/poly , dann gilt EXPTIME = Σp2.
Folgerung: Falls EXPTIME ⊆ P/poly , dann gilt P 6= NP . Beweis:
1. aus EXPTIME ⊆ P/poly folgt EXPTIME = Σp2
2. ware P = NP , so ware P = Σp2
3. ⇒ P = EXPTIME4. dies widerspricht dem Zeithierarchie-Theorem
![Page 73: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/73.jpg)
Satz von Meyer
32 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls EXPTIME ⊆ P/poly , dann gilt EXPTIME = Σp2.
Folgerung: Falls EXPTIME ⊆ P/poly , dann gilt P 6= NP . Beweis:
1. aus EXPTIME ⊆ P/poly folgt EXPTIME = Σp2
2. ware P = NP , so ware P = Σp2
3. ⇒ P = EXPTIME
4. dies widerspricht dem Zeithierarchie-Theorem
![Page 74: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/74.jpg)
Satz von Meyer
32 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls EXPTIME ⊆ P/poly , dann gilt EXPTIME = Σp2.
Folgerung: Falls EXPTIME ⊆ P/poly , dann gilt P 6= NP . Beweis:
1. aus EXPTIME ⊆ P/poly folgt EXPTIME = Σp2
2. ware P = NP , so ware P = Σp2
3. ⇒ P = EXPTIME4. dies widerspricht dem Zeithierarchie-Theorem
![Page 75: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/75.jpg)
Satz von Karp-Lipton
33 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls NP ⊆ P/poly , dann gilt PH = Σp2.
d.h. die polynomielle Hierarchie fallt zusammenes wird aber PH ) Σp
2 vermutet⇒ NP * P/poly
⇒ P 6= NP
![Page 76: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/76.jpg)
Satz von Karp-Lipton
33 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls NP ⊆ P/poly , dann gilt PH = Σp2.
d.h. die polynomielle Hierarchie fallt zusammen
es wird aber PH ) Σp2 vermutet⇒ NP * P/poly
⇒ P 6= NP
![Page 77: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/77.jpg)
Satz von Karp-Lipton
33 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls NP ⊆ P/poly , dann gilt PH = Σp2.
d.h. die polynomielle Hierarchie fallt zusammenes wird aber PH ) Σp
2 vermutet
⇒ NP * P/poly
⇒ P 6= NP
![Page 78: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/78.jpg)
Satz von Karp-Lipton
33 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls NP ⊆ P/poly , dann gilt PH = Σp2.
d.h. die polynomielle Hierarchie fallt zusammenes wird aber PH ) Σp
2 vermutet⇒ NP * P/poly
⇒ P 6= NP
![Page 79: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/79.jpg)
Satz von Karp-Lipton
33 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Falls NP ⊆ P/poly , dann gilt PH = Σp2.
d.h. die polynomielle Hierarchie fallt zusammenes wird aber PH ) Σp
2 vermutet⇒ NP * P/poly
⇒ P 6= NP
![Page 80: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/80.jpg)
schwere Funktionen
34 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Motivation: Bei TMs ist es schwierig, lowerbounds zu beweisen. Wiesieht es bei Boolean Circuits aus?
![Page 81: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/81.jpg)
schwere Funktionen
34 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Existenz schwerer Funktionen
Fur jedes n > 1 existiert eine Funktion f : {0,1}n → {0,1}, die nichtdurch einen Boolean circuit der Große 2n/(10n) berechenbar ist.(Beweis: siehe Tafel)
![Page 82: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/82.jpg)
schwere Funktionen
34 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Existenz schwerer Funktionen
Fur jedes n > 1 existiert eine Funktion f : {0,1}n → {0,1}, die nichtdurch einen Boolean circuit der Große 2n/(10n) berechenbar ist.(Beweis: siehe Tafel)
Beschreibt auch nur ein solches f eine Sprache L ∈ NP , so wareP 6= NP !
![Page 83: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/83.jpg)
schwere Funktionen
34 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Existenz schwerer Funktionen
Fur jedes n > 1 existiert eine Funktion f : {0,1}n → {0,1}, die nichtdurch einen Boolean circuit der Große 2n/(10n) berechenbar ist.(Beweis: siehe Tafel)
Beschreibt auch nur ein solches f eine Sprache L ∈ NP , so wareP 6= NP !
Bisher ist es aber nicht gelungen, das zu finden.
![Page 84: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/84.jpg)
nonuniform hierachy theorem
35 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In etwa: Haben wir einen großeren Circuit zur Verfugung, dann konnenwir damit auch mehr berechnen.
Es seien T ,T ′ : N→N Funktionen mit2n/n > T ′(n) > 10T (n) > n.Dann gilt SIZE(T (n)) ( SIZE(T ′(n))
![Page 85: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/85.jpg)
nonuniform hierachy theorem
35 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In etwa: Haben wir einen großeren Circuit zur Verfugung, dann konnenwir damit auch mehr berechnen.
Es seien T ,T ′ : N→N Funktionen mit2n/n > T ′(n) > 10T (n) > n.
Dann gilt SIZE(T (n)) ( SIZE(T ′(n))
![Page 86: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/86.jpg)
nonuniform hierachy theorem
35 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In etwa: Haben wir einen großeren Circuit zur Verfugung, dann konnenwir damit auch mehr berechnen.
Es seien T ,T ′ : N→N Funktionen mit2n/n > T ′(n) > 10T (n) > n.Dann gilt SIZE(T (n)) ( SIZE(T ′(n))
![Page 87: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/87.jpg)
Weitere Vorgehensweise
36 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition von KomplexitatsklassenEinfuhrung sinnvoller BeschrankungenWozu kann man Circuits denn nun verwenden?
Neue Ansatze zur Untersuchung von KomplexitatsklassenBessere Beschreibung paralleler Algorithmen
![Page 88: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/88.jpg)
37 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Parallele Algorithmen
![Page 89: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/89.jpg)
Parallele Algorithmen
38 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In diesem Kapitel beschaftigen wir uns mit den folgenden Fragen:Wie definiert man effiziente Parallelalgorithmen?
Welche Beziehung gibt es zur Klasse der effizienten sequenziellenAlgorithmen?Welche Beispiele gibt es fur effizient losbare Sprachen?Was hat das ganze mit Boolean circuits zu tun?
![Page 90: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/90.jpg)
Parallele Algorithmen
38 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In diesem Kapitel beschaftigen wir uns mit den folgenden Fragen:Wie definiert man effiziente Parallelalgorithmen?Welche Beziehung gibt es zur Klasse der effizienten sequenziellenAlgorithmen?
Welche Beispiele gibt es fur effizient losbare Sprachen?Was hat das ganze mit Boolean circuits zu tun?
![Page 91: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/91.jpg)
Parallele Algorithmen
38 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In diesem Kapitel beschaftigen wir uns mit den folgenden Fragen:Wie definiert man effiziente Parallelalgorithmen?Welche Beziehung gibt es zur Klasse der effizienten sequenziellenAlgorithmen?Welche Beispiele gibt es fur effizient losbare Sprachen?
Was hat das ganze mit Boolean circuits zu tun?
![Page 92: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/92.jpg)
Parallele Algorithmen
38 Simon Bischof, Philipp Christian Loewner:Boolean circuits
In diesem Kapitel beschaftigen wir uns mit den folgenden Fragen:Wie definiert man effiziente Parallelalgorithmen?Welche Beziehung gibt es zur Klasse der effizienten sequenziellenAlgorithmen?Welche Beispiele gibt es fur effizient losbare Sprachen?Was hat das ganze mit Boolean circuits zu tun?
![Page 93: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/93.jpg)
Effiziente Parallelalgorithmen
39 Simon Bischof, Philipp Christian Loewner:Boolean circuits
![Page 94: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/94.jpg)
Effiziente Parallelalgorithmen
39 Simon Bischof, Philipp Christian Loewner:Boolean circuits
poly(n)
![Page 95: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/95.jpg)
Effiziente Parallelalgorithmen
39 Simon Bischof, Philipp Christian Loewner:Boolean circuits
poly(n)
![Page 96: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/96.jpg)
Effiziente Parallelalgorithmen
39 Simon Bischof, Philipp Christian Loewner:Boolean circuits
poly(n)
![Page 97: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/97.jpg)
Effiziente Parallelalgorithmen
39 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition: Ein Problem hat einen effizienten Parallelalgorithmus wenneine Eingabe der Große n mit nO(1) Prozessoren in einer Zeit vonlogO(1)(n) gelost werden kann.
![Page 98: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/98.jpg)
Beispiele
40 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Addition zweier Zahlen mit carry lookaheadMultiplikation und Division zweier GanzzahlenMatrixoperationen: Produkt, Rang, Determinante und InverseGraphen: kurzeste Pfade und minimale Spannbaume
![Page 99: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/99.jpg)
Beispiel: carry-lookahead-Addition
41 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Bekannt und geliebt aus den TI-Vorlesungen
(Bild aus Wikipedia)
![Page 100: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/100.jpg)
Beispiel: carry-lookahead-Addition
41 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Clou hier: Vorberechnung des carry-flags
(Bild aus Wikipedia)
![Page 101: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/101.jpg)
Die Klasse NC
42 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Fur jedes d ist eine Sprache L genau dann in NCd , wennsie von einer Circuit-Familie {Cn} erkannt wird, wobeiCn eine Große von poly(n) und
eine Tiefe von O(logd n) hat undlogspace-uniform ist
NC :=⋃
i≥1NC i
Vergleich zu P/poly : NC hat starkere Anforderungen an Tiefe undUniformitat
![Page 102: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/102.jpg)
Die Klasse NC
42 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Fur jedes d ist eine Sprache L genau dann in NCd , wennsie von einer Circuit-Familie {Cn} erkannt wird, wobeiCn eine Große von poly(n) und
eine Tiefe von O(logd n) hat undlogspace-uniform ist
NC :=⋃
i≥1NC i
Vergleich zu P/poly : NC hat starkere Anforderungen an Tiefe undUniformitat
![Page 103: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/103.jpg)
Die Klasse AC
43 Simon Bischof, Philipp Christian Loewner:Boolean circuits
AC i wird ahnlich zu NC i definiertUnterschied: die UND- bzw. ODER-Gatter durfen eine unbegrenzteAnzahl Eingange habenAC :=
⋃i≥1AC i
Bemerkung: es giltNC i ⊆ AC i ⊆ NC i+1
NC = AC
![Page 104: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/104.jpg)
Die Klasse AC
43 Simon Bischof, Philipp Christian Loewner:Boolean circuits
AC i wird ahnlich zu NC i definiertUnterschied: die UND- bzw. ODER-Gatter durfen eine unbegrenzteAnzahl Eingange habenAC :=
⋃i≥1AC i
Bemerkung: es giltNC i ⊆ AC i ⊆ NC i+1
NC = AC
![Page 105: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/105.jpg)
Zur Bemerkung AC i ⊆ NC i+1
44 Simon Bischof, Philipp Christian Loewner:Boolean circuits
. . .
. . .
. . .
![Page 106: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/106.jpg)
Beispiel
45 Simon Bischof, Philipp Christian Loewner:Boolean circuits
PARITY := {x | x hat ungerade Anzahl von 1en}
PARITY ∈ NC1
Begrundung: BinarbaumWurzel des Baumes: AusgangsknotenBlatter: Eingangsknotenlinker / rechter Vorganger: Paritat des linken / rechten Teilbaumeslogarithmische Tiefe des Baumes (klar)
(Wurzel und Blatter: Umkehren der Richtung)
Bemerkung: PARITY /∈ AC0
![Page 107: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/107.jpg)
Beispiel
45 Simon Bischof, Philipp Christian Loewner:Boolean circuits
PARITY := {x | x hat ungerade Anzahl von 1en}
PARITY ∈ NC1
Begrundung: BinarbaumWurzel des Baumes: AusgangsknotenBlatter: Eingangsknotenlinker / rechter Vorganger: Paritat des linken / rechten Teilbaumeslogarithmische Tiefe des Baumes (klar)
(Wurzel und Blatter: Umkehren der Richtung)
Bemerkung: PARITY /∈ AC0
![Page 108: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/108.jpg)
Beispiel
45 Simon Bischof, Philipp Christian Loewner:Boolean circuits
PARITY := {x | x hat ungerade Anzahl von 1en}
PARITY ∈ NC1
Begrundung: BinarbaumWurzel des Baumes: AusgangsknotenBlatter: Eingangsknotenlinker / rechter Vorganger: Paritat des linken / rechten Teilbaumeslogarithmische Tiefe des Baumes (klar)
(Wurzel und Blatter: Umkehren der Richtung)
Bemerkung: PARITY /∈ AC0
![Page 109: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/109.jpg)
P-Vollstandigkeit
46 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Analog zur NP-Vollstandigkeit, aber logspace-Transformation stattpolynomieller Transformation
Beobachtungen: Sei L P-VollstandigL ∈ NC gdw. P = NCL ∈ LOGSPACE gdw. P = LOGSPACE
Beispiel: CIRCUIT-EVAL
![Page 110: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/110.jpg)
P-Vollstandigkeit
46 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Analog zur NP-Vollstandigkeit, aber logspace-Transformation stattpolynomieller Transformation
Beobachtungen: Sei L P-VollstandigL ∈ NC gdw. P = NCL ∈ LOGSPACE gdw. P = LOGSPACE
Beispiel: CIRCUIT-EVAL
![Page 111: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/111.jpg)
P-Vollstandigkeit
46 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Analog zur NP-Vollstandigkeit, aber logspace-Transformation stattpolynomieller Transformation
Beobachtungen: Sei L P-VollstandigL ∈ NC gdw. P = NCL ∈ LOGSPACE gdw. P = LOGSPACE
Beispiel: CIRCUIT-EVAL
![Page 112: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/112.jpg)
PH und EXPTIME
47 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Definition: Eine Circuit-Familie {Cn}n≥1 heißt DC-uniform wenn dieFunktionen SIZE, TYPE und EDGE in polynomieller Zeit berechenbarsind.
L ∈ PH genau dann wenn L durch eine DC-uniforme Circuit-Familieerkannt wird, wobei
1. die Große ist 2nO(1)und die Tiefe konstant
2. die UND- bzw. ODER-Gatter unbegrenzte Anzahl Eingange habenkonnen
3. alle NOT-Gatter am Eingang des Circuits sind
Wird die Forderung nach konstanter Tiefe weggelassen, erhalten wirEXPTIME.
![Page 113: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/113.jpg)
Fazit
48 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Boolean Circuits
sind interessant (hoffentlich)haben tatsachlich sinnvolle Anwendungsgebietebieten eine neue Perspektive auf das P 6= NP-Problemsind aber hinsichtlich der Losung des Problems nicht wenigerkompliziert
![Page 114: Boolean circuits: ein alternatives Rechenmodell€¦ · 1 Simon Bischof, Philipp Christian Loewner:Boolean circuits Institut fur Theoretische Informatik¨ Boolean circuits: ein alternatives](https://reader034.vdocument.in/reader034/viewer/2022042208/5eab79dc11d8f227e204dddf/html5/thumbnails/114.jpg)
49 Simon Bischof, Philipp Christian Loewner:Boolean circuits
Danke fur eure Aufmerksamkeit!