Donghua University

Download Report

Transcript Donghua University

Donghua University
Emergence of cooperation through coevolving
time scale in spatial prisoner’s dilemma
Zhihai Rong (荣智海)
[email protected]
Donghua University
2010.08@The 4th China-Europe Summer
School on Complexity Science, Shanghai
© 2002 IBM Corporation
DHU
Donghua University
Acknowledgements 
Dr. Zhi-Xi Wu
Dr. Wen-Xu Wang
Dr. Petter Holme
 Zhi-Xi Wu, Zhihai Rong & Petter Holme, Phys.Rev.E,036106,2010
 Zhihai Rong, Zhi-Xi Wu & Wen-Xu Wang, Phys.Rev.E,026101, 2010
2
DHU
Donghua University
阿豺折箭 戮力一心
阿豺有子二十人。阿豺谓曰:“
汝等各奉吾一支箭。”折之地下
。俄而命母弟慕利延曰:“汝取
一支箭折之。”慕利延折之。又
曰:“汝取十九支箭折之。”延
不能折。阿豺曰:“汝曹知否?
单者易折,众则难摧,戮力一心
,然后社稷可固!”
——《魏书•吐谷浑传》
3
DHU
Donghua University
Cooperation: the basis of human societies
Robert Boyd and Sarah Mathew, A Narrow Road to Cooperation, SCIENCE,2007
4
DHU
Donghua University
Prisoner’s dilemma (囚徒困境,PD)
Cooperator: help others at a cost to themselves.
Defector: receive the benefits without providing help.
C
C
D
D
(-2,-2) (-5,-1)
(-1,-5) (-3,-3)
Whatever opponent does, player does better by
defecting…
5
DHU
Donghua University
Some rules for evolutions cooperation
Nowak MA (2006). Five rules for the evolution of cooperation. Science
Kin selection: relative
Hamilton, J. Theor. Biol.7 (1964)
Direct reciprocity: unrelated individuals
Tit for tat(TFT): nice, punishing, forgiving, but for noise…
Axelrod & Hamilton, Science 211, (1981)
Win stay, lost shift(WSLS)
Nowak, Sigmund, Nature 364, (1993)
Indirect reciprocity: reputation
Nowak, Sigmund, Nature 437 (2005).
Network reciprocity
6
DHU
Donghua University
Spatial Game Theory
M. Nowak and R. May, Evolutionary games and spatial chaos,Nature 1992
Each player x
occupying a site on a network
playing game with neighbors and obtaining payoff: Px(t)
updating rule( replicator dynamics): select a neighbor and
learn its behavior with probability ~ f(Py(t)-Px(t))
player2
C D
C  1 0 



D  b 0 
PD :1  b  2
player1
b : the temptation to defection
7
DHU
Donghua University
Evolutionary games on graphs
G. Szabo&G. Fath, Evolutionary games on graphs, Phys. Rep. 446, 2007
Cooperator frequency fc
Game Rule
Evolutionary Rule
Replacement rule
Selection rule
Best take over replicator dynamics
W(xy) =f(Py-Px)
Random
Fermi dynamics:
Preferential
W(xy)=(1+exp(x-y/κ))-1
…
Win stay, lost shift
Memory …
8
PD,SG,SH,UG,PGG,
Rock-paper-scissors…
Structure &
property
Lattice, random graph,
small-world, scale-free…
<k>, γ, rk , CC, community
DHU
Diversity of lifetime (time
scale)University
Donghua
C.Roca, J.Cuesta, A.Sánchez (2006),Physical review letters, vol.97, pp.158701.
Z.X.Wu, Z.H.Rong, P.Holme (2009), Physical Review E, vol.80, pp.36106.
The interaction time scale — how frequently the individuals
interact with each other
The selection time scale — how frequently they modifies
their strategies
The selection time scale is slower than the interaction time
scale, the player has a finite lifetime.
Individuals local on a square lattice.
The fitness of i at t-th generation: fi(t)=afi(t-1)+(1-a)gi ,
where -- gi is the payoff of i
-- a characterizes the maternal effects.
With probability pi, an individual i is selected to update its
1
strategy: W ( s  s ) 
i
j
1  exp[( fi  f j ) /  ]
where κ characterizes the rationality of individuals, and is set as 0.01.
1/pi is the lifetime of i’s current strategy, f(0)=1.
9
DHU
Donghua University
Some key quantities to characterize
the cooperative behaviors
Frequency of cooperators: fc
player2
C D
C  1 0 
player1 


D
b
0



PD :1  b  2
The extinction threshold of
defectors/cooperators:
bc1 and bc2
10
AllD
C & D
coexist
AllC
DHU
Monomorphic time scale
Donghua University
a↗fc ↗
Optimal fc occurs
at p=0.1 for a=0.9
p1, C is frequently exploited by D.
P0, Ds around the boundary have enough time to
obtain a fitness high enough to beat Cs.
Coherence resonance
 M. Perc, New J. Phys. 2006,M. Perc & M. Marhl,New J. Phys. 2006
 J. Ren, W.-X. Wang, & F. Qi, Phys. Rev. E 75,2007
11
DHU
Polymorphic time scale
Donghua University
The leaders are the individual with low p
the followers are the individual with high p.
v% of individuals’ p are 0.1, and others’ p are 0.9.
v=0.5, a=0.9, b=1.1, fc ≈0.7
12
DHU
Donghua University
Coevolving time scale
Z.H.Rong, Z.X. Wu, W.X.Wang, Emergence of cooperation through coevolving time scale in
spatial prisoner's dilemma, submitted to Physical Review E , 82, 026101 , 2010
“win-slower, lose-faster” rule:
i updates its strategy by comparing with neighbor j with a
1
different strategy with probability W ( s  s ) 
i
j
1  exp[( fi  f j ) /  ]
If i successfully resists the invasion of j, the winner i is rewarded by
owing longer lifetime: pi=pi-β, where β is reward factor
If i accepts j's strategy, the loser i has to shorten its lifetime:
pi=pi+α, where α is punishment factor
0.1 ≤ pi≤1.0, initially pi=1.0, κ=0.01
What kind of social norm parameters (α,β) can promote the
mergence of cooperation?
13
High time scale C(p>0.5) High time scale D(p>0.5)
DHU time scale C (p≤0.5) Low time scale D(p
Donghua
Low
≤0.5) University
The extinction threshold
of cooperators, rD
a
(α, β)=(0.9,0.9)
Long-term D cluster
(α, β)=(0.9,0.1)
Long-term C cluster
(α, β)=(0.9,0.05)
short-term C cluster
14
(α, β)=(0.0,0.1)
(α, β)=(0.2,0.1)
DHU
α=0, increasing β(reward)
Donghua University
Initially p=1, pmin=0.1
 High time scale C High time scale D
 Low time scale C Low time scale D
t=100
15
t=50000
High time scale C High time scale D
DHU time scale C Low time scale D
Low
Donghua University
a
(α, β)=(0.9,0.1)
16
(α, β)=(0.0,0.1)
(α, β)=(0.2,0.1)
DHU
β =0.1, increasing α(punishment)
Donghua University
α↗, fc↗
Feedback mechanism for C/D:
 Winner Cfc↗fintess↗
 Winner Dfc↘fintess↘
α↗, their losing D neighbors
have greater chance to becoming
C, hence cooperation is promoted.
17
(α,β)=(0.1,0.1)
(α,β)=(0.9,0.1)
b=1.05
High time scale C High time scale D
DHU time scale C Low time scale D
Low
a
Donghua University
(α, β)=(0.9,0.9)
(α, β)=(0.9,0.1)
(α, β)=(0.9,0.05)
18
(α, β)=(0.0,0.1)
(α, β)=(0.2,0.1)
DHU
(α,β)=(0.9,0.9)
α =0.9, increasing β(reward)
Donghua
University
(α,β)=(0.9,0.1)
(α,β)=(0.9,0.05)
19
DHU
Donghua University
Coevolution of Teaching activity
A. Szolnoki and M. Perc, New J. Phys. 10 (2008) 043036
A. Szolnoki,et al.,Phys.Rev.E 80(2009) 021901
 The player x will adopt the randomly selected neighbor y’s strategy with:
W ( sx  s y )  wy
1
1  exp[( Px  Py ) /  ]
 wx characterizes the strength of influence (teaching activity) of x. The
leader with wx 1.
Each successful strategy adoption process is accompanied by an increase
in the donor’s teaching activity:
If y succeeds in enforcing its strategy on x, wywy+Δw.
A highly inhomogeneous distribution of influence may emerge.
20
DHU
Donghua University
Multiplicative “win-slower, lose-faster”
“win-slower, lose-faster” rule:
i updates its strategy by comparing with neighbor j with a
different strategy:
If i successfully resists the invasion of j, the winner i is rewarded
by owing longer lifetime: pi=max(pi/β, pmin)
If i accepts j's strategy, the loser i has to shorten its lifetime:
pi=min(pi*α,pmax)
pmin=0.1 and pmax=1.0
21
The extinction threshold
of cooperators, rD
DHU
Donghua University
The extinction threshold of cooperation
For loser:α↗
For winner: βmid
The additive-increase /multiplicative-decrease
(AIMD) algorithm in the TCP congestion control
on the Internet
Jacobson, Proc. ACM SIGCOMM' 88
22
The extinction threshold
of cooperators, rD
DHU
Donghua University
Conclusions
The selection time scale is slower than the
interaction time scale.
Both the fixed and the coevolving time scale.
“win-slower, lose-faster” rule
The potential application in the design of
consensus protocol in multi-agent systems.
23
DHU
Donghua University
东华大学
http://cist.dhu.edu.cn/index.asp
东华大学位于上海松江区,原名中国纺织大学,是国家
教育部所属的211全国重点大学,也是我国首批具有博士、
硕士、学士三级学位授予权的大学之一。
信息学院现有“控制理论与控制工程(90)”和“模式识别
与智能系统(02)”2个博士点以及7个硕士点,“控制科学
与工程(03)”一级学科博士后流动站,拥有“教育部数字
化纺织服装技术工程研究中心”。
信息学院现有教职工近120人,其中校特聘教授2人,长
江特聘讲座教授1人,博士生导师16人,具有正高级职称
25人,副高级职称41人。
24
DHU
Donghua University
THANKS!
Discussing
Rong Zhihai (荣智海):[email protected]
Department of Automation, DHU
25