Unsupervised Learning of Global Factors in Deep Generative Models
This work addresses the problem of capturing global dependencies in deep generative models for researchers working on unsupervised learning and disentangled representations, offering an incremental improvement over existing semi-supervised methods.
This paper introduces a deep generative model using non-i.i.d. variational autoencoders to capture global dependencies in an unsupervised manner. The model combines a mixture model in the local space with a global Gaussian latent variable, resulting in interpretable disentangled representations without explicit regularization, domain alignment capabilities, and the ability to discriminate between groups of observations.
We present a novel deep generative model based on non i.i.d. variational autoencoders that captures global dependencies among observations in a fully unsupervised fashion. In contrast to the recent semi-supervised alternatives for global modeling in deep generative models, our approach combines a mixture model in the local or data-dependent space and a global Gaussian latent variable, which lead us to obtain three particular insights. First, the induced latent global space captures interpretable disentangled representations with no user-defined regularization in the evidence lower bound (as in $β$-VAE and its generalizations). Second, we show that the model performs domain alignment to find correlations and interpolate between different databases. Finally, we study the ability of the global space to discriminate between groups of observations with non-trivial underlying structures, such as face images with shared attributes or defined sequences of digits images.