Disentangling the independently controllable factors of variation by interacting with the world

TL;DR

This paper proposes a method for discovering independently controllable factors of variation in an environment by interacting with it, aiming to improve representation learning without relying on static data or external rewards.

Contribution

It introduces a novel objective function that enables the discovery of controllable factors through interaction, advancing unsupervised disentanglement in dynamic environments.

Findings

Successfully disentangles controllable environmental factors

Operates without external reward signals

Works in interactive, dynamic settings

Abstract

It has been postulated that a good representation is one that disentangles the underlying explanatory factors of variation. However, it remains an open question what kind of training framework could potentially achieve that. Whereas most previous work focuses on the static setting (e.g., with images), we postulate that some of the causal factors could be discovered if the learner is allowed to interact with its environment. The agent can experiment with different actions and observe their effects. More specifically, we hypothesize that some of these factors correspond to aspects of the environment which are independently controllable, i.e., that there exists a policy and a learnable feature for each such aspect of the environment, such that this policy can yield changes in that feature with minimal changes to other features that explain the statistical variations in the observed data. We propose a specific objective function to find such factors, and verify experimentally that it can indeed disentangle independently controllable aspects of the environment without any extrinsic reward signal.