On Oblivious PTAS's for Nash Equilibrium

TL;DR

This paper explores the limits of oblivious PTAS algorithms for finding approximate Nash equilibria, showing both their potential in specific cases and their fundamental limitations in general games, with new algorithms for certain classes.

Contribution

It introduces new oblivious PTAS algorithms for games with specific equilibrium properties and establishes lower bounds, demonstrating the near-optimality of existing oblivious algorithms.

Findings

Oblivious PTAS exists for games with equilibria having small-O(1/n) values.

Oblivious PTAS applies to games with sparse payoff matrices, which are PPAD-complete.

Any oblivious PTAS for anonymous games with two strategies and three player types must have super-polynomial dependence on 1/eps.

Abstract

If a game has a Nash equilibrium with probability values that are either zero or Omega(1) then this equilibrium can be found exhaustively in polynomial time. Somewhat surprisingly, we show that there is a PTAS for the games whose equilibria are guaranteed to have small-O(1/n)-values, and therefore large-Omega(n)-supports. We also point out that there is a PTAS for games with sparse payoff matrices, which are known to be PPAD-complete to solve exactly. Both algorithms are of a special kind that we call oblivious: The algorithm just samples a fixed distribution on pairs of mixed strategies, and the game is only used to determine whether the sampled strategies comprise an eps-Nash equilibrium; the answer is yes with inverse polynomial probability. These results bring about the question: Is there an oblivious PTAS for Nash equilibrium in general games? We answer this question in the negative; our lower bound comes close to the quasi-polynomial upper bound of [Lipton, Markakis, Mehta 2003]. Another recent PTAS for anonymous games is also oblivious in a weaker sense appropriate for this class of games (it samples from a fixed distribution on unordered collections of mixed strategies), but its runtime is exponential in 1/eps. We prove that any oblivious PTAS for anonymous games with two strategies and three player types must have 1/eps^c in the exponent of the running time for some c>1/3, rendering the algorithm in [Daskalakis 2008] essentially optimal within oblivious algorithms. In contrast, we devise a poly(n) (1/eps)^O(log^2(1/eps)) non-oblivious PTAS for anonymous games with 2 strategies and any bounded number of player types. Our algorithm is based on the construction of a sparse (and efficiently computable) eps-cover of the set of all possible sums of n independent indicators, under the total variation distance. The size of the cover is poly(n) (1/ eps^{O(log^2 (1/eps))}.