Dynamic Latent Separation for Deep Learning

TL;DR

This paper introduces a dynamic latent separation method inspired by atomic physics that enhances model expressiveness, interpretability, and diversity without supervision, improving performance across classification and generation tasks.

Contribution

It presents a novel atom modeling approach that dynamically distances data samples in latent space, revealing sub-component importance and enabling partial interpretability.

Findings

Improves model expressiveness and interpretability.

Enhances diversity in model outputs.

Boosts performance in classification and generation tasks.

Abstract

A core problem in machine learning is to learn expressive latent variables for model prediction on complex data that involves multiple sub-components in a flexible and interpretable fashion. Here, we develop an approach that improves expressiveness, provides partial interpretation, and is not restricted to specific applications. The key idea is to dynamically distance data samples in the latent space and thus enhance the output diversity. Our dynamic latent separation method, inspired by atomic physics, relies on the jointly learned structures of each data sample, which also reveal the importance of each sub-component for distinguishing data samples. This approach, atom modeling, requires no supervision of the latent space and allows us to learn extra partially interpretable representations besides the original goal of a model. We empirically demonstrate that the algorithm also enhances the performance of small to larger-scale models in various classification and generation problems.