Natural representation of composite data with replicated autoencoders

TL;DR

This paper introduces a novel unsupervised autoencoder-based method that optimizes local entropy through replicated autoencoders, enabling robust inference of basic features in composite biological data without prior knowledge.

Contribution

It presents a new approach using replicated autoencoders and local entropy optimization for better feature inference in complex data.

Findings

Successfully infers hidden features correlating with generative processes

Enhances robustness and performance over standard autoencoders

Applicable to synthetic and protein sequence data

Abstract

Generative processes in biology and other fields often produce data that can be regarded as resulting from a composition of basic features. Here we present an unsupervised method based on autoencoders for inferring these basic features of data. The main novelty in our approach is that the training is based on the optimization of the `local entropy' rather than the standard loss, resulting in a more robust inference, and enhancing the performance on this type of data considerably. Algorithmically, this is realized by training an interacting system of replicated autoencoders. We apply this method to synthetic and protein sequence data, and show that it is able to infer a hidden representation that correlates well with the underlying generative process, without requiring any prior knowledge.