Poisson Variational Autoencoder

TL;DR

The paper introduces the Poisson VAE, a novel neural architecture that encodes sensory inputs into discrete spike counts, aligning more closely with biological neurons and improving sample efficiency in classification tasks.

Contribution

It develops the Poisson VAE combining predictive coding with discrete spike encoding, and analyzes its geometric and functional advantages over traditional VAEs.

Findings

P-VAE encodes inputs in higher dimensions enabling better class separation.

P-VAE achieves 5x sample efficiency in classification.

The model links sparse coding with metabolic cost in neural encoding.

Abstract

Variational autoencoders (VAEs) employ Bayesian inference to interpret sensory inputs, mirroring processes that occur in primate vision across both ventral (Higgins et al., 2021) and dorsal (Vafaii et al., 2023) pathways. Despite their success, traditional VAEs rely on continuous latent variables, which deviates sharply from the discrete nature of biological neurons. Here, we developed the Poisson VAE (P-VAE), a novel architecture that combines principles of predictive coding with a VAE that encodes inputs into discrete spike counts. Combining Poisson-distributed latent variables with predictive coding introduces a metabolic cost term in the model loss function, suggesting a relationship with sparse coding which we verify empirically. Additionally, we analyze the geometry of learned representations, contrasting the P-VAE to alternative VAE models. We find that the P-VAE encodes its inputs in relatively higher dimensions, facilitating linear separability of categories in a downstream classification task with a much better (5x) sample efficiency. Our work provides an interpretable computational framework to study brain-like sensory processing and paves the way for a deeper understanding of perception as an inferential process.