Learning Efficient Representation for Intrinsic Motivation

TL;DR

This paper introduces a novel deep learning method to efficiently estimate empowerment, an intrinsic motivation measure, directly from visual observations in unknown dynamics, overcoming sampling limitations of previous approaches.

Contribution

It develops a new approach using a stochastic latent space model and the Water-Filling algorithm to compute empowerment without sampling, applicable to high-dimensional visual data.

Findings

The embedding preserves information-theoretic properties of dynamics.

The method enables empowerment estimation from visual observations.

It outperforms existing sampling-based approaches in unknown environments.

Abstract

Mutual Information between agent Actions and environment States (MIAS) quantifies the influence of agent on its environment. Recently, it was found that the maximization of MIAS can be used as an intrinsic motivation for artificial agents. In literature, the term empowerment is used to represent the maximum of MIAS at a certain state. While empowerment has been shown to solve a broad range of reinforcement learning problems, its calculation in arbitrary dynamics is a challenging problem because it relies on the estimation of mutual information. Existing approaches, which rely on sampling, are limited to low dimensional spaces, because high-confidence distribution-free lower bounds for mutual information require exponential number of samples. In this work, we develop a novel approach for the estimation of empowerment in unknown dynamics from visual observation only, without the need to sample for MIAS. The core idea is to represent the relation between action sequences and future states using a stochastic dynamic model in latent space with a specific form. This allows us to efficiently compute empowerment with the "Water-Filling" algorithm from information theory. We construct this embedding with deep neural networks trained on a sophisticated objective function. Our experimental results show that the designed embedding preserves information-theoretic properties of the original dynamics.