Learning Multimodal VAEs through Mutual Supervision
This work addresses the challenge of learning from partially-observed multimodal data, which is incremental as it builds on prior VAE methods by introducing a novel combination approach.
The paper tackled the problem of modeling joint distributions over heterogeneous multimodal data and learning shared representations, by introducing MEME, a multimodal VAE that uses mutual supervision to combine information implicitly. The result showed that MEME outperforms baselines on standard metrics for both partial and complete observation schemes on MNIST-SVHN and CUB datasets.
Multimodal VAEs seek to model the joint distribution over heterogeneous data (e.g.\ vision, language), whilst also capturing a shared representation across such modalities. Prior work has typically combined information from the modalities by reconciling idiosyncratic representations directly in the recognition model through explicit products, mixtures, or other such factorisations. Here we introduce a novel alternative, the MEME, that avoids such explicit combinations by repurposing semi-supervised VAEs to combine information between modalities implicitly through mutual supervision. This formulation naturally allows learning from partially-observed data where some modalities can be entirely missing -- something that most existing approaches either cannot handle, or do so to a limited extent. We demonstrate that MEME outperforms baselines on standard metrics across both partial and complete observation schemes on the MNIST-SVHN (image-image) and CUB (image-text) datasets. We also contrast the quality of the representations learnt by mutual supervision against standard approaches and observe interesting trends in its ability to capture relatedness between data.