Game Theory and the Web

Lecturer Professor Paul Spirakis

 

Official Introduction:

The Internet and the Web are not only a huge artifact but also a major social phenomenon, dominating our everyday lives in many ways. But the Web/Internet is operated, used, managed and evolving through the interactions of many, competing or cooperating parties with self-interests.

• What science can help us to understand and affect such a system?
• What is the role of Computer Science in that in the new century?

The talk goes through a Theory of Complex Selfish Systems and tries to answer some of the above questions. We will discuss games, equilibria, evolutionary dynamics, and complexity issues that may shed some light to our understanding of the Web, the prime example of a Complex System of our times. We will argue that a new science tries to manifest itself! We hope that the talk will motivate researchers to contribute to some of the mentioned issues.

 

 

---

Basic model:

 

Game G = (N, {Si}, {Ui})

    where

N: set of players

Si: strategies of player i

Ui: XSi -> R, the utility function of player i

 

Expected utility theorem of Von Neumann & Morgenstern

 

 

Rational behaviour

Doninanat strategies (do not always exists)

Nash equilibria, Each player will not benifit if she deviates unilaterally. Every finite game has at least one Nash equilibriam. Only exponential algorithms exist now.

 

 

Discreate Math (Graphs) any directed grsph with indegress and outdegrees at most 1 must has source and sink

=>

spener lemma (combinator) any legal coloring of a triangulated pilytope contains a trichromatic trangle.

=>

fixpoint theorem of brower (analysis) Any continuous function from a polytop to itself has a fixpoint

=>

kakutanis therom => Maket equilibris

=>

Nash's theorem

 

How do we prove existence?

Pigeonhole principle

 

Approximate equilibria

best poly-time result: e = 0.34

 

 

今天来的大牛不太多，Peter，海狸鼠，PCOD的Mckee，Patrik（第一次在FPmeeting以外的seminar看到他，不过他坐我前面一直在准备后天AFP的讲义根本没听，估计也是卖人面子来的）剩下的两个年纪大的先生都没见过，剩下是一帮年轻人估计就是postDoc,Phd还有我这样的小虾米……

 

基本上说，听到亲爱的教授的口音的时候我就被打败了~~真难以想象他在美国那么多年怎么混过来的，不过至少说明他技术上肯定是很牛~~感觉自己没有能了解太多，基本的感觉是他在围绕这个Game model做应用算法方面的研究。这个研究有可能可以根据现有的状态做出有效的预测。

 

T=ST(p)/OPT

记得貌似大概解释是这样滴：in the example, assume God gonna do some change on the world and ask everyone on the earth, the result is OPT, which is the happest for everyone.In the reality, it is impossilbe, so just ask some people instead of all, the result is ST(p).  T is the stablility of the decision.还有另外一个ratio也是跟着连个值相关的，不过不记得了，当时的状况真的很窘，n多先跑路的，n多开小差的，n个打哈欠的……