Emergent Mind

Abstract

This paper presents a new variant of fictitious play (FP) called team-fictitious-play (Team-FP) that can reach equilibrium in multi-team competition, different from the other variants of FP. We specifically focus on zero-sum potential team games with network separable interactions (ZSPTGs), unifying potential games (if there is a single team) and zero-sum polymatrix games (if each team has a single member) due to their wide range of applications from robotics to financial markets beyond two-team games. Similar to the FP dynamics, in Team-FP, agents follow a simple behavioral rule where they respond (with some inertia and exploration in the update of actions) to the last actions of the neighboring team members and the beliefs formed about the other neighbors' strategies as if the opponents are playing according to some stationary strategy. We show the almost sure convergence of the empirical averages of teams' action profiles to near team-Nash equilibrium in ZSPTGs under standard assumptions on the step sizes used. We formulate a bound on the approximation error, decaying with the exploration in the agents' responses. We further examine the performance of the Team-FP dynamics numerically.

We're not able to analyze this paper right now due to high demand.

Please check back later (sorry!).

Generate a summary of this paper on our Pro plan:

We ran into a problem analyzing this paper.

Newsletter

Get summaries of trending comp sci papers delivered straight to your inbox:

Unsubscribe anytime.