Structured Disentangled Representations
This addresses a bottleneck in unsupervised representation learning for AI by enabling better control over disentangled representations, though it is incremental as it builds on prior modifications to objective functions.
The paper tackles the problem of disentangling discrete factors of variation in deep latent-variable models, which existing methods often fail to do, and demonstrates that their proposed hierarchical objective improves disentanglement and generalization to unseen factor combinations.
Deep latent-variable models learn representations of high-dimensional data in an unsupervised manner. A number of recent efforts have focused on learning representations that disentangle statistically independent axes of variation by introducing modifications to the standard objective function. These approaches generally assume a simple diagonal Gaussian prior and as a result are not able to reliably disentangle discrete factors of variation. We propose a two-level hierarchical objective to control relative degree of statistical independence between blocks of variables and individual variables within blocks. We derive this objective as a generalization of the evidence lower bound, which allows us to explicitly represent the trade-offs between mutual information between data and representation, KL divergence between representation and prior, and coverage of the support of the empirical data distribution. Experiments on a variety of datasets demonstrate that our objective can not only disentangle discrete variables, but that doing so also improves disentanglement of other variables and, importantly, generalization even to unseen combinations of factors.