Disentangled (Un)Controllable Features

TL;DR

This paper introduces a novel method for disentangling latent features in high-dimensional state spaces of MDPs into controllable and uncontrollable parts, enhancing interpretability and enabling planning within the controllable features.

Contribution

The paper proposes a new approach to disentangle features into controllable and uncontrollable partitions, improving interpretability in high-dimensional MDP representations.

Findings

Partitioned representations are interpretable across various environments.

Planning algorithms can be effectively applied within the controllable latent space.

The approach demonstrates feasibility in procedurally generated maze environments.

Abstract

In the context of MDPs with high-dimensional states, downstream tasks are predominantly applied on a compressed, low-dimensional representation of the original input space. A variety of learning objectives have therefore been used to attain useful representations. However, these representations usually lack interpretability of the different features. We present a novel approach that is able to disentangle latent features into a controllable and an uncontrollable partition. We illustrate that the resulting partitioned representations are easily interpretable on three types of environments and show that, in a distribution of procedurally generated maze environments, it is feasible to interpretably employ a planning algorithm in the isolated controllable latent partition.