Learning Multimodal VAEs Through Mutual Supervision

Joy, Tom; Shi, Yuge; Torr, Philip H.S.; Rainforth, Tom; Schmon, Sebastian M.; Siddharth, N.

Abstract

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.

Citation

Joy, T., Shi, Y., Torr, P. H., Rainforth, T., Schmon, S. M., & Siddharth, N. (2022). Learning Multimodal VAEs Through Mutual Supervision.

• Conference Name: ICLR 2022: The Tenth International Conference on Learning Representations
• Conference Location: Virtual
• Start Date: Apr 25, 2022
• End Date: Apr 29, 2022
• Acceptance Date: Jan 20, 2022
• Online Publication Date: Sep 29, 2021
• Publication Date: 2022
• Deposit Date: Jun 24, 2022
• Publicly Available Date: Jun 24, 2022
• Publisher URL: https://openreview.net/forum?id=1xXvPrAshao