Sparsely Changing Latent States for Prediction and Planning in Partially Observable Domains

TL;DR

This paper introduces GateL0RD, a recurrent neural network architecture with a bias for stable, sparsely changing latent states, improving prediction, planning, and interpretability in partially observable environments.

Contribution

GateL0RD incorporates a novel gating function and L0 penalty to promote stable, sparsely changing latent states, enhancing performance and interpretability over existing RNNs.

Findings

Outperforms state-of-the-art RNNs in prediction and control tasks.

Encodes underlying environment factors and ignores spurious dependencies.

Improves sampling efficiency and generalization in planning and reinforcement learning.

Abstract

A common approach to prediction and planning in partially observable domains is to use recurrent neural networks (RNNs), which ideally develop and maintain a latent memory about hidden, task-relevant factors. We hypothesize that many of these hidden factors in the physical world are constant over time, changing only sparsely. To study this hypothesis, we propose Gated $L_0$ Regularized Dynamics (GateL0RD), a novel recurrent architecture that incorporates the inductive bias to maintain stable, sparsely changing latent states. The bias is implemented by means of a novel internal gating function and a penalty on the $L_0$ norm of latent state changes. We demonstrate that GateL0RD can compete with or outperform state-of-the-art RNNs in a variety of partially observable prediction and control tasks. GateL0RD tends to encode the underlying generative factors of the environment, ignores spurious temporal dependencies, and generalizes better, improving sampling efficiency and overall performance in model-based planning and reinforcement learning tasks. Moreover, we show that the developing latent states can be easily interpreted, which is a step towards better explainability in RNNs.