![Page 2: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/2.jpg)
S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems”
A.Borodin, J.Boyar, K.S.Larsen“Priority Algorithms for Graph Optimization Problems ”
![Page 3: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/3.jpg)
Motivation
• Greedy algorithm are widely used in almost every kind of problem
• They are quite often the simplest way to solve a problem
• Until now there have been very few tries to analyse them as a class of algorithm
• The definition itself is quite unprecise
![Page 4: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/4.jpg)
What is greed?
![Page 5: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/5.jpg)
Oliver Stone’s Wall Street (1987)
The point is, ladies and gentleman, is that greed - for lack of a better
word - is good!
Greed is right!
Greed works!
![Page 6: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/6.jpg)
“The point is you can’t be too greedy”Donald J.Trump
![Page 7: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/7.jpg)
Greed /gri:d/noun [U]
a very strong wish to continually get more of something, especially food or money
(%om Cambridge Advance Learner’s Dictionary)
![Page 8: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/8.jpg)
The term “greedy algorithm” is didactical, elegant and
intuitively understandable, unluckily it lacks the
precision needed for a mathematical analysis.
![Page 9: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/9.jpg)
The priority model
![Page 10: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/10.jpg)
Fixed priority algorithmDetermine an allowable ordering of the set of possible input items (without knowing the actual input set of items)
While is not empty
index of input Item that comes first in the ordering
Make an irrevocable decision concerning and remove it from
S
S
S
next := I ∈ S
Inext
![Page 11: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/11.jpg)
Adaptive priority algorithm
While is not empty Determine a total ordering of all possible input
items (without knowing the input items in not yet considered) index of input Item that comes first in the ordering
Make an irrevocable decision concerning
and remove it from
S
S
next := I ∈ S
Inext
S
![Page 12: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/12.jpg)
Fixed priority algorithm:Order the vertices by the number of neighbours.Add the vertices to the DS until the graph is dominated.
Adaptive priority algorithm
The next vertex will be the one with most not yetcovered neighbours.Vertices are added to the DS until the graph is dominated.
Examples - The Dominating Set Problem
![Page 13: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/13.jpg)
Examples - The Dominating Set Problem
![Page 14: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/14.jpg)
Fixed priority algorithm
Examples - The Dominating Set Problem
![Page 15: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/15.jpg)
Fixed priority algorithm
Examples - The Dominating Set Problem
![Page 16: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/16.jpg)
Fixed priority algorithm
Examples - The Dominating Set Problem
![Page 17: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/17.jpg)
Adaptive priority algorithm
Examples - The Dominating Set Problem
![Page 18: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/18.jpg)
A general lower bound technique
Interaction between two entities Solver and Adversary, Solver tries to solve the problem applying the algorithm, Adversary tries to give Solver the worst possible instance of the problem.
Adversary must be able to provide a solution whose cost/output is used to compute the approximation ratio of Solver’s solution.
This is analogous to the competitive analysis of online algorithms.
![Page 19: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/19.jpg)
Approximation ratio
An algorithm is said to have an approximation ratio of if the expected cost of the solution is within a factor of of the cost of an optimal solution, i.e. if
ρ
C ρ
C∗
C
C∗≤ ρ
holds for every instance of the problem.
![Page 20: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/20.jpg)
Given a directed Graph and two nodesfind a directed tree of edges, rooted at and with as a leaf. The objective is to minimize the combined weight of theedges on the path from to .
Shortest Path
s, t ∈ V
s
s
t
G(V, E)t
![Page 21: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/21.jpg)
Theorem
No fixed priority algorithm can solve the shortestpath problem with any approximation ratio ρ
Proof idea: a
y=1
!!!!!
!!!!
!
s
v=1
""""""""""
u=k##
x=1
$$!!!
!!!!
!
w=k%%
t
b
z=1
&&########
k ≥ 2ρ
y ≺ z
![Page 22: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/22.jpg)
1. If Solver rejects y then Adversary remove z and builds the following instance:
a
y=1
!!!
!!
!!
!!
!
s
u=k""
x=1
##!
!!
!!
!!
! t
b
Solver can no longer find a solution whereas Adversarypropose S={u,y} and wins.
![Page 23: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/23.jpg)
2. If Solver accepts y then Adversary builds the following instance:
a
y=1
!!!!!
!!!!
!
s
u=k""
x=1
##!!!
!!!!
! t
b
z=1
$$""""""""
Adversary propose S={x,z} with cost 2, Solver can’t propose {x,z,y} because it wouldn’t be a directed tree rooted at s, hence its solution must contain u and therefore it will cost at least k+1.
approximation ratio = k + 1
2>
k
2= ρ
QED
![Page 24: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/24.jpg)
As Dijkstra’s Algorithm can solve the Shortest Path Problem exactly and it belongs to the class of ADAPTIVE priority algorithms, we can conclude that the classes of algorithms FIXED and ADAPTIVE are not equivalent.
![Page 25: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/25.jpg)
Given a directed Graph in which every vertex has an associated positive weight find a vertexcover (a subset of V whose nodes touch every edge of the graph) of minimum weight.
The weight of a vertex cover is defined as:
Weighted Vertex Cover
v ∈ V w(v)
V′
w(V ′) :=∑
v∈V ′
w(v)
G(V, E)
![Page 26: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/26.jpg)
It has been shown that is not possible to approximate the weighted vertex cover with an approximation ratio better than
10√
5 − 21 = 1.3606
Unless P=NP
The best known (non priority) algorithm approximates the weighted vertex cover with an approximation ratio of
2 − θ(1
√log n
)
![Page 27: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/27.jpg)
There are quite simple adaptive priority algorithm which solves the weighted vertex cover problem with an approximation ratio of 2.
There’s for instance Clarkson’s Algorithm that at any iteration picks the node with the minimum weight on number of not yet covered edges ratio.
![Page 28: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/28.jpg)
Theorem
No adaptive priority algorithm can achieve anapproximation ratio better than 2 for the WeightedVertex Cover Problem.
![Page 29: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/29.jpg)
Proof idea:
Kn,n bipartite graph
Nodes can have a weigh)of either or1 n
2
![Page 30: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/30.jpg)
One of the following events will eventually occur
1. the solver accepts a nodewith weight n
2
2. the solver accepts nodes of weightfrom either sides of the bipartite graph
n − 1
1
3. the solver rejects a node
![Page 31: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/31.jpg)
case 1 the solver accepts a node vwith weight n
2
1
1
n2
*
![Page 32: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/32.jpg)
Adversary set all the node on the opposite side to and the remaining node to
1
n2
1
1
n2
1
1
1
1
n2
n2
n2
*
case 1 the solver accepts a node vwith weight n
2
![Page 33: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/33.jpg)
Adversary set all the node on the opposite side to and the remaining node to
1
n2
1
1
n2
1
1
1
1
n2
n2
n2
*Solver’s solution contains vand therefore cost at least n
2
case 1 the solver accepts a node vwith weight n
2
![Page 34: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/34.jpg)
Adversary set all the node on the opposite side to and the remaining node to
1
n2
1
1
n2
1
1
1
1
n2
n2
n2
*Solver’s solution contains vand therefore cost at leastn
2
Adversary proposes a solutionthat costs n
case 1 the solver accepts a node vwith weight n
2
![Page 35: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/35.jpg)
Adversary set all the node on the opposite side to and the remaining node to
1
n2
1
1
n2
1
1
1
1
n2
n2
n2
*Solver’s solution contains vand therefore cost at leastn
2
ρ < n2/n = nIf Adversarywins, with this is alwaysthe case
ρ < 2
Adversary proposes a solutionthat costs n
case 1 the solver accepts a node vwith weight n
2
![Page 36: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/36.jpg)
case 2 the solver accepts nodes of weightfrom either sides of the bipartite graph
1
1
n − 1 1
1
1
![Page 37: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/37.jpg)
1
1
1
1 Adversary sets the last nodeon “Solver’s” side to and all the other nodes to
n2
1
n2
1
1
1
1
1
n − 1 1case 2 the solver accepts nodes of weightfrom either sides of the bipartite graph
![Page 38: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/38.jpg)
Solver’s solution either contains the “heavy” node or all the nodeson the other side, hence it willcost at least 2n − 1
n − 1 1case 2 the solver accepts nodes of weightfrom either sides of the bipartite graph
1
1
1
1
n2
1
1
1
1
1
Adversary sets the last nodeon “Solver’s” side to and all the other nodes to
n2
1
![Page 39: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/39.jpg)
Solver’s solution either contains the “heavy” node or all the nodeson the other side, hence it willcost at least 2n − 1
Adversary’s solution costs n
IfAdversary wins
ρ <2n − 1
n= 2 − o(1)
case 2 the solver accepts nodes of weightfrom either sides of the bipartite graph
n − 1 1
1
1
1
1
n2
1
1
1
1
1
Adversary sets the last nodeon “Solver’s” side to and all the other nodes to
n2
1
![Page 40: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/40.jpg)
1
1
*
case 3 the solver rejects a node v (of any weight)
![Page 41: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/41.jpg)
Adversary set all the unseen node on the opposite side of v to and the remaining node to
case 3 the solver rejects a node v (of any weight)
1
1
*
n2
n2
n2
n2
n2
1
1
1
1
![Page 42: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/42.jpg)
case 3 the solver rejects a node v (of any weight)
1
1
*
n2
n2
n2
n2
1
1
1
At least 2 nodes are set to n2
Adversary set all the unseen node on the opposite side of v to and the remaining node to
n2
1
![Page 43: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/43.jpg)
case 3 the solver rejects a node v (of any weight)
1
1
*
n2
n2
n2
n2
1
1
1
At least 2 nodes are set to n2
Solver’s solution doesn’t contain*, hence it cost at least 2n
2
Adversary set all the unseen node on the opposite side of v to and the remaining node to
n2
1
![Page 44: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/44.jpg)
case 3 the solver rejects a node v (of any weight)
1
1
*
n2
n2
n2
n2
1
1
1
At least 2 nodes are set to n2
2n2
Adversary’s solution contains vand costs at most n2
+ n − 1
Adversary wins iff ρ <2n2
n2 + n − 1= 2 − o(1)
QED
n2
1
Adversary set all the unseen node on the opposite side of v to and the remaining node to
Solver’s solution doesn’t contain*, hence it cost at least
![Page 45: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/45.jpg)
so far so good!
![Page 46: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/46.jpg)
Acceptances-first algorithm• Special kind of adaptive priority algorithm
• The decision is an accept/reject decision
• After the first rejection the algorithm can no longer accept any other item
![Page 47: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/47.jpg)
Memoryless• Special kind of adaptive priority algorithm
• The decision is an accept/reject decision
• Rejections have no influx on the further decision (i.e. rejections are seen as no-ops)
![Page 48: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/48.jpg)
Memoryless algorithm can be simulated by acceptances-firs) algorithms
![Page 49: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/49.jpg)
Node vs. edge model• In the node model the Graph is represented by
lists of adjacent vertices.
• In the edge model the Graph is represented by lists of adjacent edges.
• The two models should be equivalent as they represent the same thing
• unfortunately most results in the two papers require the problem to be formulated in a specified form
![Page 50: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/50.jpg)
Conclusions• This approach leads to results that hold for whole
classes of algorithms including yet to be designed algorithms
• The distinction between edge model and node model is inelegant
• Not all the result are really significant
• The model is promising but still in the alpha version
![Page 51: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/51.jpg)
dulcis in fundocourtesy by our friend at MSN.COM
![Page 52: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/52.jpg)
Say you want to go from Haugesund to Trondheim
Approximatively 600 km
![Page 53: Greedy algorithms on graphs - DISCOElia Noris norisel@student.ethz.ch S.Davis, R.Impagliazzo “Models of greedy Algorithms for Graph Problems” A.Borodin, J.Boyar, K.S.Larsen “Priority](https://reader033.vdocument.in/reader033/viewer/2022060921/60aca8b5fa0d0b52c5266186/html5/thumbnails/53.jpg)
It doesn’t matter if you’re looking for the shortest way...
or maybe for the fastest...
the point is....
WHERE DO YOU WANT TO GO TODAY?