Learning Disentangled Representations with Semi-Supervised Deep Generative Models

TL;DR

This paper introduces a semi-supervised deep generative model framework based on variational autoencoders that learns disentangled data representations by combining strong assumptions on some variables with neural network flexibility.

Contribution

It proposes a novel model architecture and objective for semi-supervised learning of disentangled representations, extending standard VAEs with a flexible graphical model structure.

Findings

Successfully learns interpretable, disentangled features

Demonstrates improved generative and discriminative performance

Applicable across various datasets and model configurations

Abstract

Variational autoencoders (VAEs) learn representations of data by jointly training a probabilistic encoder and decoder network. Typically these models encode all features of the data into a single variable. Here we are interested in learning disentangled representations that encode distinct aspects of the data into separate variables. We propose to learn such representations using model architectures that generalise from standard VAEs, employing a general graphical model structure in the encoder and decoder. This allows us to train partially-specified models that make relatively strong assumptions about a subset of interpretable variables and rely on the flexibility of neural networks to learn representations for the remaining variables. We further define a general objective for semi-supervised learning in this model class, which can be approximated using an importance sampling procedure. We evaluate our framework's ability to learn disentangled representations, both by qualitative exploration of its generative capacity, and quantitative evaluation of its discriminative ability on a variety of models and datasets.