Understanding the Eluder Dimension

TL;DR

This paper investigates the eluder dimension, a complexity measure for function classes in reinforcement learning, revealing its relationship with generalized rank, and providing bounds and characterizations for specific cases.

Contribution

It establishes new bounds relating eluder dimension to generalized rank and characterizes it for binary functions, advancing understanding of complexity measures in learning theory.

Findings

Eluder dimension can be exponentially smaller than generalized rank for certain functions.

The derivative condition on activation functions is necessary for bounding eluder dimension.

Characterization of eluder dimension for binary functions in terms of star number and threshold dimension.

Abstract

We provide new insights on eluder dimension, a complexity measure that has been extensively used to bound the regret of algorithms for online bandits and reinforcement learning with function approximation. First, we study the relationship between the eluder dimension for a function class and a generalized notion of rank, defined for any monotone "activation" $\sigma : \mathbb{R}\to \mathbb{R}$, which corresponds to the minimal dimension required to represent the class as a generalized linear model. It is known that when $\sigma$ has derivatives bounded away from $0$, $\sigma$-rank gives rise to an upper bound on eluder dimension for any function class; we show however that eluder dimension can be exponentially smaller than $\sigma$-rank. We also show that the condition on the derivative is necessary; namely, when $\sigma$ is the $\mathsf{relu}$ activation, the eluder dimension can be exponentially larger than $\sigma$-rank. For binary-valued function classes, we obtain a characterization of the eluder dimension in terms of star number and threshold dimension, quantities which are relevant in active learning and online learning respectively.