one-way flow nash networks
DESCRIPTION
One-Way Flow Nash Networks. Frank Thuijsman Joint work with Jean Derks, Jeroen Kuipers, Martijn Tennekes. Outline. The Model of One-Way Flow Networks An Existence Result A Counterexample. Outline. The Model of One-Way Flow Networks An Existence Result A Counterexample. Outline. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/1.jpg)
International Conference on Modeling, Computation and Optimization New Delhi, January 9-10, 2008
One-Way Flow Nash Networks
Frank Thuijsman
Joint work with Jean Derks, Jeroen Kuipers, Martijn Tennekes
![Page 2: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/2.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20082
Outline
• The Model of One-Way Flow Networks• An Existence Result• A Counterexample
![Page 3: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/3.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20083
Outline
• The Model of One-Way Flow Networks• An Existence Result• A Counterexample
![Page 4: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/4.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20084
Outline
• The Model of One-Way Flow Networks• An Existence• A Counterexample
![Page 5: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/5.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20085
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)
![Page 6: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/6.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20086
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
![Page 7: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/7.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20087
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
![Page 8: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/8.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20088
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
• cij is the cost for agent i for making a link to agent j
![Page 9: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/9.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 20089
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
• cij is the cost for agent i for making a link to agent j
Example of a network g
2
●
1●
3
●
●4
●5
6
●
![Page 10: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/10.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200810
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
• cij is the cost for agent i for making a link to agent j
Example of a network g
Agent 1 is connected to agents 3,4,5 and 6and obtains profits v13 , v14 , v15 , v16 2
●
1●
3
●
●4
●5
6
●
![Page 11: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/11.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200811
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
• cij is the cost for agent i for making a link to agent j
Example of a network g
Agent 1 is connected to agents 3,4,5 and 6and obtains profits v13 , v14 , v15 , v16
Agent 1 is not connected to agent 2
The Model of One-Way Flow Networks
2
●
1●
3
●
●4
●5
6
●
![Page 12: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/12.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200812
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
• cij is the cost for agent i for making a link to agent j
Example of a network g
Agent 1 is connected to agents 3,4,5 and 6and obtains profits v13 , v14 , v15 , v16
Agent 1 is not connected to agent 2Agent 1 has to pay c13 for the link (3,1)
2
●
1●
3
●
●4
●5
6
●
![Page 13: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/13.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200813
The Model of One-Way Flow Networks
Network Formation Game (N,v,c)• N={1,2,3,…,n}
• vij is the profit for agent i for being connected to agent j
• cij is the cost for agent i for making a link to agent j
Example of a network g
The payoff π1(g) for agent 1 in network g is
π1(g) = v13 + v14 + v15 + v16 - c13 2
●
1●
3
●
●4
●5
6
●
![Page 14: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/14.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200814
The Model of One-Way Flow Networks
More generallyπi(g) = ∑j Є Ni(g) vij– ∑j Є Di(g) cij
where Ni(g) is the set of agents that i is connected to, and
where Di(g) is the set of agents that i is directly connected to.
2
●
1●
3
●
●4
●5
6
●
![Page 15: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/15.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200815
The Model of One-Way Flow Networks
More generallyπi(g) = ∑j Є Ni(g) vij– ∑j Є Di(g) cij
where Ni(g) is the set of agents that i is connected to, and
where Di(g) is the set of agents that i is directly connected to.
An action for agent i is any subset S of N\{i}indicating the set of agents that i connects to directly
2
●
1●
3
●
●4
●5
6
●
![Page 16: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/16.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200816
The Model of One-Way Flow Networks
More generallyπi(g) = ∑j Є Ni(g) vij– ∑j Є Di(g) cij
where Ni(g) is the set of agents that i is connected to, and
where Di(g) is the set of agents that i is directly connected to.
An action for agent i is any subset S of N\{i}indicating the set of agents that i connects to directly
A network g is a Nash networkif each agent i is playing a best responsein terms of his individual payoff πi(g)
2
●
1●
3
●
●4
●5
6
●
![Page 17: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/17.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200817
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
2
●
1●
3
●
●4
●5
6
●g
![Page 18: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/18.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200818
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
Here g-i denotes the network derived from g
by removing all direct links of agent i
2
●
1●
3
●
●4
●5
6
●g
![Page 19: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/19.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200819
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
Here g-i denotes the network derived from g
by removing all direct links of agent i
2
●
1●
3
●
●4
●5
6
●g
2
●
1●
3
●
●4
●5
6
●
![Page 20: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/20.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200820
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
Here g-i denotes the network derived from g
by removing all direct links of agent i
2
●
1●
3
●
●4
●5
6
●g
2
●
1●
3
●
●4
●5
6
●g-3
![Page 21: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/21.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200821
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
A set S that maximizes the right-hand side of above expressionis called a best response for agent i to the network g
![Page 22: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/22.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200822
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
A set S that maximizes the right-hand side of above expressionis called a best response for agent i to the network g
In a Nash network all agents are linked to their best responses
![Page 23: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/23.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200823
A Closer Look at Nash Networks
A network g is a Nash network if for each agent i πi(g) ≥ πi(g-i + {(j,i): j Є S})
for all subsets S of N\{i}
A set S that maximizes the right-hand side of above expressionis called a best response for agent i to the network g
In a Nash network all agents are linked to their best responses
If cik > ∑j≠i vij for all agents k≠i,
then the only best response for agent i is the empty set Ф
![Page 24: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/24.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200824
Homogeneous Costs
For each agent i all links are equally expensive: cij = ci for all j
![Page 25: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/25.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200825
Homogeneous Costs
For each agent i all links are equally expensive: cij = ci for all j
Observation for Homogeneous CostsIf link (j,k) exists in g,then for agent i ≠ j,k, linking with kis at least as good as linking with j
i●
k
●
●j
g
![Page 26: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/26.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200826
Homogeneous Costs
For each agent i all links are equally expensive: cij = ci for all j
Observation for Homogeneous CostsIf link (j,k) exists in g,then for agent i ≠ j,k, linking with kis at least as good as linking with j
i●
k
●
●j
g
![Page 27: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/27.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200827
Homogeneous Costs
For each agent i all links are equally expensive: cij = ci for all j
Observation for Homogeneous CostsIf link (j,k) exists in g,then for agent i ≠ j,k, linking with kis at least as good as linking with j
“Downstream Efficiency”i●
k
●
●j
g
![Page 28: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/28.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200828
Lemma
For any network formation game (N,v,c) with homogeneous costs and with ci ≤ ∑j≠i vij for all agents i,
all cycle networks are Nash networks
2
●
1●
3
●
●4
●5
6
●
![Page 29: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/29.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200829
Lemma
For any network formation game (N,v,c) with homogeneous costs and with ci ≤ ∑j≠i vij for all agents i,
all cycle networks are Nash networks
2
●
1●
3
●
●4
●5
6
●
![Page 30: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/30.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200830
Lemma
For any network formation game (N,v,c) with homogeneous costs and with ci ≤ ∑j≠i vij for all agents i,
all cycle networks are Nash networks
When removing (2,1) agent 1 loosesprofits from agents 2, 3, 4, 5, 6
2
●
1●
3
●
●4
●5
6
●
![Page 31: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/31.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200831
Lemma
For any network formation game (N,v,c) with homogeneous costs and with ci ≤ ∑j≠i vij for all agents i,
all cycle networks are Nash networks
When adding (4,1) agent 1 pays
an additional cost of c14
2
●
1●
3
●
●4
●5
6
●
![Page 32: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/32.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200832
Lemma
For any network formation game (N,v,c) with homogeneous costs and with ci ≤ ∑j≠i vij for all agents i,
all cycle networks are Nash networks
When replacing (2,1) by (4,1) agent 1looses profits from agents 2 and 3
2
●
1●
3
●
●4
●5
6
●
![Page 33: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/33.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200833
Theorem
For any network formation game (N,v,c) with homogeneous costs, a Nash network exists
![Page 34: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/34.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200834
Theorem
For any network formation game (N,v,c) with homogeneous costs, a Nash network exists
Proof by induction to the number of agents:
![Page 35: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/35.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200835
Theorem
For any network formation game (N,v,c) with homogeneous costs, a Nash network exists
Proof by induction to the number of agents:
If n=1, then the trivial network is a Nash network
![Page 36: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/36.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200836
Theorem
For any network formation game (N,v,c) with homogeneous costs, a Nash network exists
Proof by induction to the number of agents:
If n=1, then the trivial network is a Nash network
Induction hypothesis: Nash networks exist for all network games with less than n agents.
![Page 37: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/37.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200837
Theorem
For any network formation game (N,v,c) with homogeneous costs, a Nash network exists
Proof by induction to the number of agents:
If n=1, then the trivial network is a Nash network
Induction hypothesis: Nash networks exist for all network games with less than n agents.
Suppose that (N,v,c) is a network game with n agents for which NO Nash network exists.
![Page 38: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/38.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200838
Recall the Lemma:
For any network formation game (N,v,c) with homogeneous costs and with ci ≤ ∑j≠i vij for all agents i,
all cycle networks are Nash networks
2
●
1●
3
●
●4
●5
6
●
![Page 39: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/39.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200839
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
![Page 40: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/40.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200840
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent n
![Page 41: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/41.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200841
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’
![Page 42: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/42.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200842
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis)
![Page 43: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/43.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200843
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis) Then by assumption g’ is no Nash network in (N,v,c)
![Page 44: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/44.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200844
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis) Then by assumption g’ is no Nash network in (N,v,c)Therefore there is an agent ifor whom the links in g are no best response
![Page 45: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/45.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200845
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis) Then by assumption g’ is no Nash network in (N,v,c)Therefore there is an agent ifor whom the links in g are no best responseThis agent i can not be agent nso w.l.o.g. this agent is agent 1
![Page 46: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/46.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200846
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis) Then by assumption g’ is no Nash network in (N,v,c)Therefore there is an agent ifor whom the links in g are no best responseThis agent i can not be agent nso w.l.o.g. this agent is agent 1 and he has a best response T with n Є T
![Page 47: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/47.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200847
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis) Then by assumption g’ is no Nash network in (N,v,c)Therefore there is an agent ifor whom the links in g are no best responseThis agent i can not be agent nso w.l.o.g. this agent is agent 1 and he has a best response T with n Є Tand therefore c1 ≤ vin
![Page 48: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/48.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200848
Proof continued:
Hence there is at least one agent i with ci > ∑j≠i vij
W.l.o.g. this agent is agent nConsider (N’,v’,c’) with N’=N\{n} and with v and c restricted to agents in N’Let g’ be a Nash network in (N’,v’,c’) (induction hypothesis) Then by assumption g’ is no Nash network in (N,v,c)Therefore there is an agent ifor whom the links in g are no best responseThis agent i can not be agent nW.l.o.g. this agent is agent 1 and he has a best response T with n Є Tand therefore c1 ≤ vin
1
●
●n
![Page 49: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/49.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200849
Proof continued:
Now recall that for any other agent i linking to agent 1 would be at least as good as linking to agent n
vij for j≠1
Define vij* = vi1 + vin for i≠1, j=1
v11 + v1n for i=1
1
●
●ni●
![Page 50: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/50.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200850
Proof continued:
Now recall that for any other agent i linking to agent 1 would be at least as good as linking to agent n
vij for j≠1
Define vij* = vi1 + vin for i≠1, j=1
v11 + v1n for i=1
Now π*i(g) = πi(g + (n,1))
for any network g on N’ and for any agent i in N’
1
●
●ni●
![Page 51: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/51.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200851
Proof continued:
Now recall that for any other agent i linking to agent 1 would be at least as good as linking to agent n
vij for j≠1
Define vij* = vi1 + vin for i≠1, j=1
v11 + v1n for i=1
Now π*i(g) = πi(g + (n,1))
for any network g on N’ and for any agent i in N’
The game (N’,v*,c’) has a Nash network g*
1
●
●ni●
![Page 52: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/52.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200852
Proof continued:
This network g* can not be a Nash network in (N,v,c)
![Page 53: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/53.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200853
Proof continued:
This network g* can not be a Nash network in (N,v,c)Hence at least one agent is not playing a best response
![Page 54: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/54.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200854
Proof continued:
This network g* can not be a Nash network in (N,v,c)Hence at least one agent is not playing a best response
However, any agent improving in (N,v,c)contradicts that g* is a Nash network in (N’,v*,c’)by the way that v* and v are related to eachother
![Page 55: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/55.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200855
Proof continued:
This network g* can not be a Nash network in (N,v,c)Hence at least one agent is not playing a best response
However, any agent improving in (N,v,c)contradicts that g* is a Nash network in (N’,v*,c’)by the way that v* and v are related to eachother
![Page 56: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/56.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200856
Example
For network formation games (N,v,c) with heterogeneous costs, Nash networks do not need to exist
![Page 57: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/57.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200857
Example
For network formation games (N,v,c) with heterogeneous costs, Nash networks do not need to exist
A heterogeneous costs structure1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
![Page 58: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/58.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200858
Example
For network formation games (N,v,c) with heterogeneous costs, Nash networks do not need to exist
A heterogeneous costs structure
other links to agent 1 cost 1+ε
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
![Page 59: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/59.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200859
Example
For network formation games (N,v,c) with heterogeneous costs, Nash networks do not need to exist
A heterogeneous costs structure
other links to agent 1 cost 1+εother links to agent 2 cost 2+ε
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
![Page 60: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/60.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200860
Example
For network formation games (N,v,c) with heterogeneous costs, Nash networks do not need to exist
A heterogeneous costs structure
other links to agent 1 cost 1+εother links to agent 2 cost 2+εother links to agents 3 and 4 cost 3+ε
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
![Page 61: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/61.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200861
Example
For network formation games (N,v,c) with heterogeneous costs, Nash networks do not need to exist
A heterogeneous costs structure
other links to agent 1 cost 1+εother links to agent 2 cost 2+εother links to agents 3 and 4 cost 3+ε
profits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
![Page 62: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/62.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200862
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part A)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.
![Page 63: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/63.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200863
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part A)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays {2},then agent 1 plays {4}.
![Page 64: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/64.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200864
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part A)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays {2},then agent 1 plays {4}.Then agent 2 plays {1}, because agent 3 never plays {1}.
![Page 65: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/65.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200865
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part A)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays {2},then agent 1 plays {4}.Then agent 2 plays {1}, because agent 3 never plays {1}.Then agent 3 plays {2}.
![Page 66: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/66.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200866
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part A)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays {2},then agent 1 plays {4}.Then agent 2 plays {1}, because agent 3 never plays {1}.Then agent 3 plays {2}.Then agent 4 should play Ф.
![Page 67: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/67.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200867
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part A)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays {2},then agent 1 plays {4}.Then agent 2 plays {1}, because agent 3 never plays {1}.Then agent 3 plays {2}.Then agent 4 should play Ф. A contradiction
![Page 68: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/68.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200868
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.
![Page 69: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/69.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200869
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays Ф, then agent 1 plays S containing 4.
![Page 70: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/70.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200870
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays Ф, then agent 1 plays S containing 4.Then agent 2 plays {1}.
![Page 71: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/71.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200871
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays Ф, then agent 1 plays S containing 4.Then agent 2 plays {1}. Then agent 3 plays {2}.
![Page 72: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/72.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200872
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays Ф, then agent 1 plays S containing 4.Then agent 2 plays {1}. Then agent 3 plays {2}.Then agent 1 plays {3,4}.
![Page 73: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/73.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200873
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays Ф, then agent 1 plays S containing 4.Then agent 2 plays {1}. Then agent 3 plays {2}.Then agent 1 plays {3,4}. Then agent 4 should play {2}.
![Page 74: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/74.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200874
Example Explained
The cost/payoff structure
other links to 1 cost 1+εother links to 2 cost 2+εother links to 3, 4 cost 3+εprofits vij =1 for all i and j
1
●
2
●
4
●●3
1-ε
2-ε
3-ε3-ε
The arguments (part B)
In any Nash network agent 3 and agent 4 would either play {2} or Ф.If agent 4 plays Ф, then agent 1 plays S containing 4.Then agent 2 plays {1}. Then agent 3 plays {2}.Then agent 1 plays {3,4}. Then agent 4 should play {2}. Again a contradiction
![Page 75: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/75.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200875
Concluding remarks
Our proof implies that for homogeneous costs Nash networks exist that contain at most one cycleand where every vertex has outdegree at most 1
![Page 76: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/76.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200876
Concluding remarks
Our proof implies that for homogeneous costs Nash networks exist that contain at most one cycleand where every vertex has outdegree at most 1
Our model is based mainly on:• V. Bala & S. Goyal (2000): A non-cooperative model of network
formation. Econometrica 68, 1181-1229.• A. Galeotti (2006): One-way flow networks: the role of
heterogeneity. Economic Theory 29, 163-179.
![Page 77: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/77.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200877
Time for Questions
a preprint of our paper is available upon request
![Page 78: One-Way Flow Nash Networks](https://reader035.vdocuments.site/reader035/viewer/2022062422/568140ae550346895dac6d3b/html5/thumbnails/78.jpg)
International Conference on Modeling, Computation and Optimization
New Delhi, January 9-10, 200878
Thank you for your attention!
a preprint of our paper is available upon request