Multi-Channel Causal Variational Autoencoder

In this work we propose Multi-Channel Causal VAE (MC2VAE), a causal disentanglement approach for multichannel data, whose objective is to jointly learn modality-specific latent representations from a multichannel dataset, and identify a linear causal structure between the latent variables. We formally derive MC2VAE and the optimization strategy for its parameters. Experiments on synthetically generated data-sets underline the ability of our model to identify ground-truth hidden causal relationships.

Download paper here

Recommended citation: Al-Ali S., Balelli I. (2024). Multi-Channel Causal Variational Autoencoder. Preprint. [hal-04666466]