Discretization of continuous input spaces in the hippocampal autoencoder

TL;DR

This paper demonstrates that sparse autoencoder neurons can discretize continuous sensory spaces into high-dimensional, minimal-overlap codes, supporting spatial and auditory cognition and enhancing reinforcement learning performance.

Contribution

It introduces a novel approach where hippocampal-like autoencoder neurons discretize sensory spaces, unifying spatial and episodic memory functions in a computational model.

Findings

Neurons produce spatial tuning similar to place cells.

High-dimensional codes enable minimal-overlap discretization.

Reinforcement learning agents improve performance using these representations.

Abstract

The hippocampus has been associated with both spatial cognition and episodic memory formation, but integrating these functions into a unified framework remains challenging. Here, we demonstrate that forming discrete memories of visual events in sparse autoencoder neurons can produce spatial tuning similar to hippocampal place cells. We then show that the resulting very high-dimensional code enables neurons to discretize and tile the underlying image space with minimal overlap. Additionally, we extend our results to the auditory domain, showing that neurons similarly tile the frequency space in an experience-dependent manner. Lastly, we show that reinforcement learning agents can effectively perform various visuo-spatial cognitive tasks using these sparse, very high-dimensional representations.