cooperative/coalitional game theory

57
Cooperative/coalitional game theory A composite of slides taken from Vincent Conitzer and Giovanni Neglia (Modified by Vicki Allan) 1

Upload: adonai

Post on 24-Feb-2016

51 views

Category:

Documents


0 download

DESCRIPTION

Cooperative/coalitional game theory. A composite of slides taken from Vincent Conitzer and Giovanni Neglia (Modified by Vicki Allan). Prisoners Dilemma rules. Binding agreements are not possible. Note in Prisoners dilemma, if binding agreements were possible, there would be no dilemma - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Cooperative/coalitional game theory

Cooperative/coalitional game theory

A composite of slides taken from Vincent Conitzer and Giovanni Neglia

(Modified by Vicki Allan)

1

Page 2: Cooperative/coalitional game theory

Prisoners Dilemma rules

1. Binding agreements are not possible. Note in Prisoners dilemma, if binding agreements were possible, there would be no dilemma

2. Utility is given directly to individuals as a result of individual actions. So, I am not worried about collective utility.

Solution: cooperative game theory

2

Page 3: Cooperative/coalitional game theory

Cooperative Games

• Coalitions – set of agents• Grand coalition – all agents work together• Characteristic function: v:2Ag R subsets of agents are assigned a value.• Simple coalitional game: a coalition has value 0

or 1. A voting system can be thought of as a simple game (have enough votes to win or do not)

3

Page 4: Cooperative/coalitional game theory

Three Parts of Cooperative Action

1. Coalition Structure Generation – who should work together. Know only characteristic function. Want to earn a lot, but so does everyone else.

2. Optimization Problem – given coalitions and tasks, which coalition should do each problem.

3. Dividing the utility- What is fair? What is stable?

4

Page 5: Cooperative/coalitional game theory

In Multi-Agent Systems

• Agents create their own coalitions rather than have a centralized decision.

5

Page 6: Cooperative/coalitional game theory

Scenario 1 – Bargain Buy(supply-demand)

• Store “Bargain Buy” advertises a great price

• 300 people show up• 5 in stock• Everyone sees the advertised

price, but it just isn’t possible for all to achieve it

Page 7: Cooperative/coalitional game theory

Scenario 2 – hiring a new PhD(strategy)

• Universities ranked 1,2,3• Students ranked a,b,cDilemma for second tier

university• offer to “a” student• likely rejected• delay for acceptance• “b” students are gone

Page 8: Cooperative/coalitional game theory

Scenario 3 – selecting a spouse(agency)

• Bob knows all the characteristics of the perfect wife

• Bob seeks out such a wife• Why would the perfect

woman want Bob?

Page 9: Cooperative/coalitional game theory

Scenario 4 Blowhard (trust)What if one person talks a good story, but his claims of skills are really inflated?

He isn’t capable of performing. the task.

Page 10: Cooperative/coalitional game theory

Scenario 5 Power in diversityYou consult local traffic statistics to find a good route home from work

But so does everyone else

Page 11: Cooperative/coalitional game theory

The coalition is completed and rewards are earned. How are they fairly divided among agents with various contributions?If organizer is greedy, why wouldn’t others replace him with a cheaper agent?

Page 12: Cooperative/coalitional game theory

Consider the Stable Marriage Problem – small coalition

• In a small town there are n men and n women who wish to be wed. Each person would be happy to be married to any of the people of the opposite sex but has a definite preference ranking of the possible marriage partners.

• If marriages are arranged arbitrarily, some of the marriages can be unstable in the following sense. Suppose Alice marries Bob and Carol marries Dave. If Alice prefers Dave to Bob and Dave prefers Alice to Carol, then Alice and Dave will leave their partners; in this situation, we say that the marriages of Alice and Dave are unstable.

• Goal is stability not just the most partners: the difference between local and global utility.

12

Page 13: Cooperative/coalitional game theory

Core• Ex: software engineering group project. Who would you

choose to work with? Who would also choose you? Coalition doesn’t form unless everyone is happy with it.

• Stability is necessary but not sufficient for a coalition to form. If it isn’t stable, someone will defect.

If there are multiple stable coalitions, another may form instead.

• core: set of feasible distributions of payoff to members of the grand coalition so none will defect.

• We require the outcome (payments) to be both feasible (able to pay) and efficient (all utility distributed).

• Pareto efficiency is the efficiency referred to.

13

Page 14: Cooperative/coalitional game theory

• We say a coalition objects to (or blocks) an outcome if every member of the coalition is strictly better off in some other feasible outcome.

• Asking if the grand coalition is stable means, “Is the core non-empty?”

• The point of the core is to study stability not fairness.

14

Page 15: Cooperative/coalitional game theory

Example• v{a} = 1 v{b} = 3 v{a,b} = 5• How should we divide the profit if they work together?

Which are feasible, fair, efficient, stable?• <a=5,b=0>• <a=4,b=1>• <a=3,b=2>• <a=2,b=3>• <a=1,b=4>• <a=1.5,b=3.5>• <a=0,b=5>• <a=2,b=2>• <a=2,b=4>

15

Page 16: Cooperative/coalitional game theory

So what are concerns?

• What if there are no outcomes in the core?• What if there are multiple outcomes in the

core, how do you pick?• Looking at all possible distributions of utility is

exponential as you have 2n possible objecting subsets to consider.

16

Page 17: Cooperative/coalitional game theory

Example• v{1} = 2 v{2} = 2 v{3} = 2 • v{1,2} = 5 v{2,3} = 5 v{1,2,3} = 6• How should we divide the profit if they work

together? Which are feasible, fair, efficient, stable?

• <2,2,2>• <3,3,0>• <1,2,3>• <1,2.5,2.5>• <2,2.5, 1.5>

17

Page 18: Cooperative/coalitional game theory

More Examples Emptiness & multiplicity

18

• Example 1: Let us modify the above example so that agents receive no additional utility from being all together (and being alone gives 0)– v({1, 2, 3}) = 6, – v({1, 2}) = v({1, 3}) = v({2, 3}) = 6,– v({1}) = v({2}) = v({3}) = 0

Page 19: Cooperative/coalitional game theory

More Examples Emptiness & multiplicity

19

• Example 1: Let us modify the above example so that agents receive no additional utility from being all together (and being alone gives 0)– v({1, 2, 3}) = 6, – v({1, 2}) = v({1, 3}) = v({2, 3}) = 6,– v({1}) = v({2}) = v({3}) = 0

• Now the core is empty! Notice, the core must involve the grand coalition (giving payoff for each).

Page 20: Cooperative/coalitional game theory

More Examples Emptiness & multiplicity

20

• Example 2:– v({1, 2, 3}) = 18, – v({1, 2}) = v({1, 3}) = v({2, 3}) = 10,– v({1}) = v({2}) = v({3}) = 0

Page 21: Cooperative/coalitional game theory

More Examples Emptiness & multiplicity

21

• Example 2:– v({1, 2, 3}) = 18, – v({1, 2}) = v({1, 3}) = v({2, 3}) = 10,– v({1}) = v({2}) = v({3}) = 0

• Now lots of outcomes are in the core – (6, 6, 6), (5, 5, 8), …

Page 22: Cooperative/coalitional game theory

Fairness• Could we get people to join a coalition by

saying, “Will you agree to a fair solution even though it is not in the core?”

• In examplev{1} =0 v{2} = 0 v{3} = 0 v{1,2} = 5 v{2,3} = 5 v{1,2,3} = 6What would a fair but unstable solution be?• If we had extra (that no one could really

demand), how would we divide any surplus?

22

Page 23: Cooperative/coalitional game theory

Terms• Marginal Contribution (value added): μi(C) = v(C U {i}) – v(C)

• shi is what i is given (its share)• Symmetry: Agents that make the same contribution should get the

same payoff• Dummy player – never increases the value of the coalition beyond

what it could earn alone μi(C) = v(C U {i}) – v(C) = v{i} for every C

• Alternate definition of Dummy: μi(C) = 0 for every C• Additivity: if we add two games defined by v and w by letting (v+w)(S)

= v(S) + w(S), then the utility for an agent in v+w should be the sum of her utilities in v and w• most controversial axiom (for example, participant i’s cost-share of a runway

and terminal is it’s cost-share of the runway plus his cost-share of the terminal)

23

Page 24: Cooperative/coalitional game theory

Superadditivity

24

• v is superadditive if for all coalitions A, B with A∩B = Ø, v(AUB) ≥ v(A) + v(B)

• Informally, the union of two coalitions can always act as if they were separate, so should be able to get at least what they would get if they were separate.

• There is a synergy – if not, coalitions make no sense.• Usually makes sense• Previous examples were all superadditive• Given this, always efficient for grand coalition to form• Without superadditivity, finding a core is not possible.

Page 25: Cooperative/coalitional game theory

The Shapley value [Shapley 1953]

25

• In dividing the profit, sometimes agent is given its marginal contribution (how much better the group is by its addition)

• The marginal contribution scheme is unfair because it depends on the ordering of the agents

• One way to make it fair: average over all possible orderings• Let MC(i, π) be the marginal contribution of i in ordering π• Then i’s Shapley value is ΣπMC(i, π)/(n!)• The Shapley value is always in the core for convex games• … but not in general, even when core is nonempty, e.g.

– v({1, 2, 3}) = v({1, 2}) = v({1, 3}) = 1,– v = 0 everywhere else

Page 26: Cooperative/coalitional game theory

Example: v({1, 2, 3}) = v({1, 2}) = v({1, 3}) = 1,v = 0 everywhere else

1 2 3123

132

213

231

312

321avg

26

Compute the Shapley value for each.

Is the solution in the core?

Page 27: Cooperative/coalitional game theory

Axiomatic characterization of the Shapley value

27

• The Shapley value is the unique solution concept that satisfies:– (Pareto) Efficiency: the total utility is the value of the grand

coalition, Σi in Nu(i) = v(N)– Symmetry: two symmetric players (add the same amount to

coalitions they join) must receive the same utility– Dummy: if v(S {i}) = v(S) for all S, then i must get 0– Additivity: if we add two games defined by v and w by

letting (v+w)(S) = v(S) + w(S), then the utility for an agent in v+w should be the sum of her utilities in v and w.

Page 28: Cooperative/coalitional game theory

Additivity Example• Game a

28

Coalition Value 1 2 31 0 123 0 5 12 0 132 0 6 03 0 213 5 0 1

12 5 231 6 0 013 0 312 0 6 023 0 312 6 0 0

123 6 17/6 17/6 2/6

Coalition Value 1 2 31 0 123 0 5 02 0 132 0 0 53 0 213 5 0 0

12 5 231 0 0 513 5 312 5 0 023 5 312 0 5 0

123 5 10/6 10/6 10/6

• Game b

Page 29: Cooperative/coalitional game theory

a+b

29

Coalition Value 1 2 3

1 0 123 0 10 1

2 0 132 0 6 5

3 0 213 10 0 1

12 10 231 6 0 5

13 5 312 5 6 0

23 5 312 6 5 0

123 11 27/6 27/6 12/6

Notice how their value in the combined game is the sum of their values in the original games

Page 30: Cooperative/coalitional game theory

Shapley Result• Satisfies all three fairness axioms (the last

three listed)• If you came up with some other method that

satisfied all fairness axioms, it would be Shapley. In other words, the Shapley value is the UNIQUE value that satisfies all fairness axioms.

30

Page 31: Cooperative/coalitional game theory

Computing a solution in the core

31

• How do you even represent the characteristic function which defines the problem? exponential in the number of agents.

• Can use linear programming:– Variables: u(i)– Distribution constraint: Σi in Nu(i) = v(N)– Non-blocking constraints: for every S, Σi in Su(i) ≥ v(S)

• Problem: number of constraints exponential in number of players (as you have values for all possible subsets)

• … but is this practical?

Page 32: Cooperative/coalitional game theory

Convexity

32

• v is convex if for all coalitions A, B, v(AUB)-v(B) ≥ v(A)-v(A∩B)

• In other words, the amount A adds to B (in forming the union) is at least as much it adds to the intersection.

• One interpretation: the marginal contribution of an agent is increasing in the size of the set that it is added to. The term marginal contribution means the additional contribution. Precisely, the marginal contribution of A to B is v(AUB)-v(B)

• Example, suppose we have three independent researchers. When we combine them at the same university, the value added by A is greater if the set is larger.

Page 33: Cooperative/coalitional game theory

Example• v{1} = 0 v{2} = 0 v{3} = 0 • v{1,2} = 8 v{2,3} = 8 v{1,2,3} = 12

• Let A = {1,2} and B={2,3}

• v(AUB)-v(B) ≥ v(A)-v(A∩B)

33

Page 34: Cooperative/coalitional game theory

Convexity

34

• In convex games, core is always nonempty. (Core doesn’t require convexity, but convexity produces a core.)

• One easy-to-compute solution in the core: agent i gets u(i) = v({1, 2, …, i}) - v({1, 2, …, i-1})– Marginal contribution scheme- each agent is rewarded by what it adds

to the union.– Works for any ordering of the agents

Page 35: Cooperative/coalitional game theory

Theory of cooperative games with sidepayments

• It starts with von Neumann and Morgenstern (1944)

• Two main (related) questions:– which coalitions should form?– how should a coalition (which forms) divide its

winnings among its members?• The specific strategy the coalition will

follow is not of particular concern...• Note: there are also cooperative games

without sidepayments35

Page 36: Cooperative/coalitional game theory

Example: Minimum Spanning Tree game

• For some games the characteristic form representation is immediate

• Communities 1,2 & 3 want to be connected to a nearby power source– Possible transmission links & costs as in figure

36

source

1

2

3100

40

5020

4040

Page 37: Cooperative/coalitional game theory

Example: Minimum Spanning Tree game

• Communities 1,2 & 3 want to be connected to a nearby power source

37

v(void) = 0v(1) = 0v(2) = 0v(3) = 0v(12) = -90 + 100 + 50 = 60v(13) = -80 + 100 + 40 = 60v(23) = -60 + 50+ 40 = 30v(123) = -100 + 100 + 50+ 40 = 90

A strategically equivalent game.We show what is gained from the

coalition. How to divide the gain?

source

1

2

3100

40

5020

4040

Page 38: Cooperative/coalitional game theory

The Core• What about MST game? We use value to mean

what is saved by going with a group.– v(void)= v(1) = v(2) = v(3)=0– v(12) = 60, v(13) = 60, v(23) = 30– v(123) = 90

• Analitically, in getting to a group of three, you must make sure you do better than the group of 2 cases:– x1+x2>=60, iff x3<=30– x1+x3>=60, iff x2<=30– x2+x3>=30, iff x1<=60

381 2

Page 39: Cooperative/coalitional game theory

The Core• Let’s choose an imputation in the core:

x=(60,25,5)• The payoffs represent the savings, the

costs under x are – c(1)=100-60=40, – c(2)=50-25=25– c(3)=40-5=35

39

source

1

2

3100

40

5020

4040

FAIR?

Page 40: Cooperative/coalitional game theory

The Shapley value: computation• MST game

– v(void) = v(1) = v(2) = v(3)=0– v(1,2) = 60, v(1,3) = 60, v(2,3) = 30, v(1,2,3) = 90

40

1 2 3123132213231312321avg

Value added by

Coa

litio

ns

Page 41: Cooperative/coalitional game theory

The Shapley value: computation• MST game

– v(void) = v(1) = v(2) = v(3)=0– v(12) = 60, v(13) = 60, v(23) = 30, v(123) = 90

41

1 2 3123 0 60 30

132213231312321avg

Value added by

Coa

litio

ns

Page 42: Cooperative/coalitional game theory

The Shapley value: computation

• MST game– v(void) = v(1) = v(2) = v(3)=0– v(12) = 60, v(13) = 60, v(23) = 30, v(123) = 90

1 2 3123 0 60 30

132 0 30 60

213 60 0 30

231 60 0 30

312 60 30 0

321 60 30 0

avg 40 25 25

Value added by

Coa

litio

ns

Discountper person

Page 43: Cooperative/coalitional game theory

The Shapley value: computation

• A faster way• The amount player i contributes to

coalition S, of size s, is v(S)-v(S-i)• This contribution occurs for those

orderings in which i is preceded by the s-1 other players in S, and followed by the n-s players not in S

• ki = 1/n! S:i in S (s-1)! (n-s)! (v(S)-v(S-i))

43

Page 44: Cooperative/coalitional game theory

The Shapley value has been used for cost sharing. Suppose threeplanes share a runway. The planes require 1, 2, and 3 KM to land. Let’s label those planes 1, 2, and 3.Thus, a runway of 3 must be build, but how much should each pay?

Instead of looking at utility given, look at how much increased cost was required.

1 2 3123132213231312321avg

Page 45: Cooperative/coalitional game theory

The Shapley value has been used for cost sharing. Suppose threeplanes share a runway. The planesrequire 1, 2, and 3 KM to land.Thus, a runway of 3 must bebuild, but how much should each pay?

Instead of looking at utility given, look at how much increased cost was required.

Page 46: Cooperative/coalitional game theory

An application: voting power• A voting game is a pair (N,W) where N is the set of players

(voters) and W is the collection of winning coalitions, s.t.– the empty set is not in W (it is a losing coalition) – N is in W (the coalition of all voters is winning)– if S is in W and S is a subset of T then T is in W

• Also weighted voting game can be considered in which each player has w voting weight an a quota is required to win.

• The Shapley value of a voting game is a measure of voting power (Shapley-Shubik power index)– The winning coalitions have payoff 1– The losing coalitions have payoff 0

46

Page 47: Cooperative/coalitional game theory

All powers sum to 1. What is power?

• {99,99,1: 100} Any two win, so all have equal power 1/3.

• {6,4,2: 10} Having non-zero weight does not guarantee power. 2 has zero power. It is a dummy.

• If the same agent exists in all winning coalitions, the core is non-empty (as no subgroup can pull out and win).

• Simple game: any coalition is either winning (1) or losing (0)

• It seems clear that some simple games could not be formulated in these terms. (incomplete)

47

Page 48: Cooperative/coalitional game theory

An application: voting power• The United Nations Security Council in 1954

– 5 permanent members (P)– 6 non-permanent members (N)– the winning coalitions had to have at least 7 members,– but the permanent members had veto power

• A winning coalition had to have at least seven members including all the permanent members

• The seventh member joining the coalition is the pivotal one: he makes the coalition winning

• Is this a fair voting system?

48

Page 49: Cooperative/coalitional game theory

An application: voting power• 11! possible orderings• Power of non permanent members

– (PPPPPN)N(NNNN)– 5 ways to pick which N comes before and 4! possible arrangements for those

that come after. (5! total choices)– 6! possible arrangements for (PPPPPN)– The total number of arrangements in which an N is pivotal is 6!5!– The power of all non permanent members is 6!5!/11!, each getting .0022

• The power of each permanent members is.1974• The ratio of power of a P member to a N member is 91:2• In 1965

– 5 permanent members (P)– 10 non-permanent members (N)– the winning coalitions has to have at least 9 members,– the permanent members keep the veto power

• Similar calculations lead to a ratio of power of a P member to a N member equal to 105:1 (.19627:.001865)

49

Page 50: Cooperative/coalitional game theory

How to represent the characteristic function (in condensed way)?

• Induced subgraphsNode represents agents. Arc represents added

benefit if source/target agents are both in the same coalition

50

Page 51: Cooperative/coalitional game theory

Example

51

V(ab) = 3v(ac) = 2v(abc) = 5v(bd) = 6v(d) = 5v(cd) = 9

v(abc) = 7 How would we change the diagram?

Page 52: Cooperative/coalitional game theory

Example

52

What do you think of this method? Any concerns?

Enforces superadditivity.

Clearly, in linear space, we can’t expect to represent an exponentially rich set of rewards.

What if B added to either A or C adds 3, but the gain isn’t experienced twice?

Do we need to worry about cases where B adds 3 to A, but if C is also present, the gain is actually negative?

Page 53: Cooperative/coalitional game theory

We say the method is …

• succinct – as it is much easier to represent• incomplete – as there are characteristic functions

for which the method doesn’t work• How would we use the weighted graph to

compute Shapley value? It seems like each node should get half the weight of the arc it shares.

• We can prove that by breaking in into small problems and using Shapley’s additivity axiom.

53

Page 54: Cooperative/coalitional game theory

Marginal Contribution Nets

• Extension of idea of induced subgraphs where total weight is sum of weight due to parts.

• We have a series of patterns (boolean expressions involving agents) and values. To determine the characteristic value for a subset, you add up the values (positive or negative) associated with the rule.

54

Page 55: Cooperative/coalitional game theory

Example

#1: a Λ b 5 marginal increase#2: b 2#3: c 4#4: b Λ ¬c -2 so

55

Possible with induced graphs, right?

Coalition Value Rules

a 0noneb 0#2,#4c 4#3ab 5#1,#2,#4ac 4#3bc 6#2. #3abc 11#1,#2,#3

Page 56: Cooperative/coalitional game theory

Thought questions?• Are the rules powerful enough? According to researchers, all characteristic

functions can be represented, but we aren’t told how to generate the rules!!!

• Plus, is it really any simpler? Instead of showing all combinations, you now have all the original combinations PLUS the combinations with negative components.

56

Page 57: Cooperative/coalitional game theory

Majority game

• A group of agents S has a value of 1 if |S| > n/2.

• Can we do this with Induced subgraph?• Can we do this with marginal contribution

nets?

57