27.01.2014 Views

Multiagent Systems

Multiagent Systems

Multiagent Systems

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

of. Dr. Jürgen Dix · Department of Informatics, TUC <strong>Multiagent</strong> <strong>Systems</strong>, WS 06/07 402/731<br />

5. Nets and coalitions 4. Payoff Division<br />

Theorem 5.31 (Shapley-Value)<br />

There is only one payoff division satisfying the above three<br />

axioms. It is called the Shapley value of agent i and is<br />

defined by<br />

x i = 1<br />

|A|!<br />

∑<br />

(|A| − |S| − 1)!|S|!(v S∪{i} − v S )<br />

S⊆A<br />

(|A| − S)! is the number of all possible joining orders of the<br />

agents (to form a coalition).<br />

(v S∪{i} − v S ) is i’s marginal contribution when added to set S.<br />

There are |S|! ways for S to be built before i’s joining. There are<br />

(|A| − |S| − 1)! ways for the remaining agents to form S (after i).<br />

The Shapley value sums up the marginal contributions of agent i<br />

averaged over all joining orders.<br />

An expected gain can be computed by taking a random joining<br />

order and computing the Shapley value.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!