47
Strategic Network Formation and Group Formation Elliot Anshelevich Rensselaer Polytechnic Institute (RPI)

Strategic Network Formation and Group Formation

  • Upload
    rowa

  • View
    34

  • Download
    0

Embed Size (px)

DESCRIPTION

Strategic Network Formation and Group Formation. Elliot Anshelevich Rensselaer Polytechnic Institute (RPI). Centralized Control. A majority of network research has made the centralized control assumption: - PowerPoint PPT Presentation

Citation preview

Page 1: Strategic Network Formation                      and Group Formation

Strategic Network Formation and Group Formation

Elliot Anshelevich

Rensselaer Polytechnic Institute (RPI)

Page 2: Strategic Network Formation                      and Group Formation

Centralized Control

A majority of network research has made the centralized control assumption:

Everything acts according to a centrally defined and specified algorithm

This assumption does not make sense in many cases.

Page 3: Strategic Network Formation                      and Group Formation

Self-Interested Agents

• Internet is not centrally controlled• Many other settings have self-interested agents• To understand these, cannot assume centralized control

• Algorithmic Game Theory studies such networks

Page 4: Strategic Network Formation                      and Group Formation

Agents in Network Design

• Traditional network design problems are centrally controlled

• What if network is instead built by many self-interested agents?

• Properties of resulting network may be very different from the globally optimum one

s

Page 5: Strategic Network Formation                      and Group Formation

Goal

• Compare networks created by self-interested agents with the optimal network– optimal = cheapest

– networks created by self-interested agents = Nash equilibria

• Can realize any Nash equilibrium by finding it, and suggesting it to players– Requires central coordination

– Does not require central control

OPT

NE

s

Page 6: Strategic Network Formation                      and Group Formation

The Price of Stability

Price of Anarchy = cost(worst NE)

cost(OPT)

Price of Stability = cost(best NE)

cost(OPT)

[Koutsoupias, Papadimitriou]s

t1…tk

1 k

Can think of latter as a network designer proposing a solution.

Page 7: Strategic Network Formation                      and Group Formation

Single-Source Connection Game[A, Dasgupta, Tardos, Wexler 2003]

Given: G = (V,E), k terminal nodes, costs ce for all e E

Each player wants to build a network in which his node is connected to s.

Each player selects a path, pays for some portion of edges in path (depends on cost sharing scheme)

s

Goal: minimize payments,while fulfilling connectivity requirements

Page 8: Strategic Network Formation                      and Group Formation

Other Connectivity Requirements

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation

[A, Caskurlu 2009]

[A, Dasgupta, Tardos, Wexler 2003]

Page 9: Strategic Network Formation                      and Group Formation

Group Network Formation Games

Terminal Backup: Each terminal wants to connect to k other terminals.

Page 10: Strategic Network Formation                      and Group Formation

Group Network Formation Games

“Group Steiner Tree”: Each terminal wants to connect to at least one terminal from each color.

Terminal Backup: Each terminal wants to connect to k other terminals.

Page 11: Strategic Network Formation                      and Group Formation

Other Connectivity Requirements

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation: every agent wants to connect to a group

that provides enough resources

satisfactory group specified by a monotone set function

[A, Caskurlu 2009]

[A, Dasgupta, Tardos, Wexler 2003]

[A, Caskurlu 2009]

Page 12: Strategic Network Formation                      and Group Formation

Centralized Optimum

Single-source Connection Game: Steiner Tree.

Sets of nodes: Steiner Forest.

Survivable: Generalized Steiner Forest.

Terminal Backup: Cheapest network where each terminal connected to at least k other terminals.

“Group Steiner Tree”: Cheapest where every component is a Group Steiner Tree.

Corresponds to constrained forest problems, has 2-approx.

Page 13: Strategic Network Formation                      and Group Formation

Connection Games

Given: G = (V,E), k players, costs ce for all e E

Each player wants to build a network where his connectivity requirements are satisfied.

Each player selects subgraph, pays for some portion of edges in it (depends on cost sharing scheme)

s

Goal: minimize payments,while fulfilling connectivity requirements

NE

Page 14: Strategic Network Formation                      and Group Formation

Sharing Edge Costs

How should multiple players

on a single edge split costs?

One approach: no restrictions...

...any division of cost agreed upon by players is OK. [ADTW 2003, HK 2005, EFM 2007, H 2009, AC 2009]

Another approach: try to ensure some sort of fairness.

[ADKTWR 2004, CCLNO 2006, HR 2006, FKLOS 2006]

Page 15: Strategic Network Formation                      and Group Formation

Connection Games with Fair Sharing

Given: G = (V,E), k players, costs ce for all e E

Each player selects subnetwork where his connectivity requirements are satisfied.

Players using e pay for it evenly: ci(P) = Σ ce/ke

( ke = # players using e )

s

Goal: minimize payments,while fulfilling connectivity requirements

e є P

Page 16: Strategic Network Formation                      and Group Formation

Fair Sharing

Fair sharing: The cost of each edge e is shared equally by the users of e

Advantages:

• Fair way of sharing the cost

• Nash equilibrium exists

• Price of Stability is at most log(# players)

Page 17: Strategic Network Formation                      and Group Formation

Price of Stability with Fairness

Price of Anarchy is large

Price of Stability is at most log(# players)

Proof: This is a Potential Game, so Nash equilibrium exists Best Response converges Can use this to show existence of good equilibrium

s

t1…tk

1 k

Page 18: Strategic Network Formation                      and Group Formation

Fair Sharing

Fair sharing: The cost of each edge e is shared equally by the users of e

Advantages:

• Fair way of sharing the cost

• Nash equilibrium exists

• Price of Stability is at most log(# players)

Disadvantages:

• Player payments are constrained, need to enforce fairness

• Price of stability can be at least log(# players)

Page 19: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

Demands:1-t, 2-t, 3-t

Page 20: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

Minimum Cost Solution (of cost 1+)

Page 21: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

Each player chooses a path P.Cost to player i is:

cost(i) =

(Everyone shares cost equally)

cost(P)# using P

Page 22: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

Player 3 pays (1+ε)/3,

could pay 1/3

Page 23: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

so player 3

would deviate

Page 24: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

now player 2

pays (1+ε)/2,

could pay 1/2

Page 25: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

so player 2

deviates also

Page 26: Strategic Network Formation                      and Group Formation

Example: Self-Interested Behavior

1 12

13

1 2 3

t

0 0 0

1+

Player 1 deviates as well, giving a solution with cost 1.833.

This solution is stable/ this solution is a Nash Equilibrium.

It differs from the optimal solution by a factor of 1+ + Hk = Θ(log k)!

1 12 3

Page 27: Strategic Network Formation                      and Group Formation

Sharing Edge Costs

How should multiple players

on a single edge split costs?

One approach: no restrictions...

...any division of cost agreed upon by players is OK. [ADTW 2003, HK 2005, EFM 2007, H 2009, AC 2009]

Another approach: try to ensure some sort of fairness.

[ADKTWR 2004, CCLNO 2006, HR 2006, FKLOS 2006]

Page 28: Strategic Network Formation                      and Group Formation

Example: Unrestricted Sharing

Fair Sharing: differs from the optimal solution by a factor of Hk = Θ(log k)

Unrestricted Sharing: OPT is a stable solution

1 12

13

1 2 3

t

0 0 0

1+

Page 29: Strategic Network Formation                      and Group Formation

Contrast of Sharing Schemes

Unrestricted Sharing Fair Sharing

NE don’t always exist NE always exist

P.o.S. = O(k) P.o.S. = O(log(k))

(P.o.S. = Price of Stability)

Page 30: Strategic Network Formation                      and Group Formation

Contrast of Sharing Schemes

Unrestricted Sharing Fair Sharing

NE don’t always exist NE always exist

P.o.S. = O(k) P.o.S. = O(log(k))

P.o.S. = 1 for P.o.S. = (log(k)) for

many games almost all games

(P.o.S. = Price of Stability)

Page 31: Strategic Network Formation                      and Group Formation

Contrast of Sharing Schemes

Unrestricted Sharing Fair Sharing

NE don’t always exist NE always exist

P.o.S. = O(k) P.o.S. = O(log(k))

P.o.S. = 1 for P.o.S. = (log(k)) for

many games almost all games

OPT is an approx. NE OPT may be far from NE

(P.o.S. = Price of Stability)

Page 32: Strategic Network Formation                      and Group Formation

Unrestricted Sharing Model

What is a NE in this model?

• Player i picks payments for each edge e. (strategy = vector of payments)

• Edge e is bought if total payments for it ≥ ce.

• Any player can use bought edges.

Page 33: Strategic Network Formation                      and Group Formation

Unrestricted Sharing Model

• Player i picks payments for each edge e. (strategy = vector of payments)

• Edge e is bought if total payments for it ≥ ce.

• Any player can use bought edges.

What is a NE in this model?

Payments so that no players want to change them

Page 34: Strategic Network Formation                      and Group Formation

Unrestricted Sharing Model

• Player i picks payments for each edge e. (strategy = vector of payments)

• Edge e is bought if total payments for it ≥ ce.

• Any player can use bought edges.

What is a NE in this model?

Payments so that no players want to change them

Page 35: Strategic Network Formation                      and Group Formation

Connection Games with Unrestricted Sharing

Given: G = (V,E), k players, costs ce for all e E

Strategy: a vector of payments

Players choose how much to pay, buy edges together

s

Goal: minimize payments,while fulfilling connectivity requirements

Cost(v) = if v does not satisfy connectivity requirementsPayments of v otherwise

Page 36: Strategic Network Formation                      and Group Formation

Connectivity Requirements

Single-source: connect to s

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation: every agent wants to connect to a group

that provides enough resources

satisfactory group specified by a monotone set function

Page 37: Strategic Network Formation                      and Group Formation

Some Results

Single-source: connect to s

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation: every agent wants to connect to a group

that provides enough resources

satisfactory group specified by a monotone set function

OPT is a Nash Equilibrium (Price of Stability=1)

If k=n

If k=n

Page 38: Strategic Network Formation                      and Group Formation

Some Results

Single-source: connect to s

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation: every agent wants to connect to a group

that provides enough resources

satisfactory group specified by a monotone set function

OPT is a -approximate Nash Equilibrium(no one can gain more than factor by switching)

=2

=2

=3

=1

Page 39: Strategic Network Formation                      and Group Formation

Some Results

Single-source: connect to s

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation: every agent wants to connect to a group

that provides enough resources

satisfactory group specified by a monotone set function

If we pay for 1-1/ fraction of OPT, then the players will pay for the rest

=2

=2

=3

=1

Page 40: Strategic Network Formation                      and Group Formation

Some Results

Single-source: connect to s

Survivable: connect to s with two disjoint paths

Sets of nodes: agent i wants to connect set Ti

Group formation: every agent wants to connect to a group

that provides enough resources

satisfactory group specified by a monotone set function

Can compute cheap approximate equilibria in poly-time

Page 41: Strategic Network Formation                      and Group Formation

Contrast of Sharing Schemes

Unrestricted Sharing Fair Sharing

NE don’t always exist NE always exist

P.o.S. = O(k) P.o.S. = O(log(k))

P.o.S. = 1 for P.o.S. = (log(k)) for

many games almost all games

OPT is an approx. NE OPT may be far from NE

(P.o.S. = Price of Stability)

Page 42: Strategic Network Formation                      and Group Formation

Contrast of Sharing Schemes

Unrestricted Sharing Fair Sharing

NE don’t always exist NE always exist

P.o.S. = O(k) P.o.S. = O(log(k))

P.o.S. = 1 for P.o.S. = (log(k)) for

many games almost all games

OPT is an approx. NE OPT may be far from NE

(P.o.S. = Price of Stability)

Page 43: Strategic Network Formation                      and Group Formation

Contrast of Sharing Schemes

Unrestricted Sharing Fair Sharing

NE don’t always exist NE always exist

P.o.S. = O(k) P.o.S. = O(log(k))

P.o.S. = 1 for P.o.S. = (log(k)) for

many games almost all games

OPT is an approx. NE OPT may be far from NE

If we really care about efficiency:Allow the players more freedom!

Page 44: Strategic Network Formation                      and Group Formation

Example: Unrestricted Sharing

Fair Sharing: differs from the optimal solution by a factor of Hk log k

Unrestricted Sharing: OPT is a stable solution

Every player gives what they can afford

1 12

13

1 2 3

t

0 0 0

1+

Page 45: Strategic Network Formation                      and Group Formation

General Techniques

To prove that OPT is an exact/approximate equilibrium:

Construct a payment scheme

Pay in order: laminar system of witness sets

If cannot pay, form deviations to create cheaper solution

'T

)( ii p

u

eT

eT

e

ip

Page 46: Strategic Network Formation                      and Group Formation

Network Destruction Games

• Each player wants to protect itself from untrusted nodes

• Have cut requirements: must be disconnected from set Ti

• Cutting edges costs money

• Can show similar results for:

Multiway Cut, Multicut, etc.

Page 47: Strategic Network Formation                      and Group Formation

Thank you.