Decontextualized learning for interpretable hierarchical representations of visual patterns

TL;DR

This paper introduces DHRL, a novel algorithm for learning interpretable, hierarchical, and disentangled features from natural images, addressing dataset limitations and enabling scientific analysis.

Contribution

The paper presents DHRL, a new training paradigm combining generative models and ladder networks for hierarchical feature learning in small datasets.

Findings

DHRL effectively learns interpretable hierarchical features.

The method is generative and compatible with empirical and theoretical analysis.

Application to evolutionary biology demonstrates practical utility.

Abstract

Apart from discriminative models for classification and object detection tasks, the application of deep convolutional neural networks to basic research utilizing natural imaging data has been somewhat limited; particularly in cases where a set of interpretable features for downstream analysis is needed, a key requirement for many scientific investigations. We present an algorithm and training paradigm designed specifically to address this: decontextualized hierarchical representation learning (DHRL). By combining a generative model chaining procedure with a ladder network architecture and latent space regularization for inference, DHRL address the limitations of small datasets and encourages a disentangled set of hierarchically organized features. In addition to providing a tractable path for analyzing complex hierarchal patterns using variation inference, this approach is generative and can be directly combined with empirical and theoretical approaches. To highlight the extensibility and usefulness of DHRL, we demonstrate this method in application to a question from evolutionary biology.