On PAC-Bayesian reconstruction guarantees for VAEs
This work addresses the problem of theoretical guarantees for VAEs, which is incremental as it builds on recent efforts to understand VAE behavior.
The paper tackles the theoretical understanding of variational autoencoders (VAEs) by analyzing their reconstruction ability for unseen test data using PAC-Bayes theory, providing generalization bounds and insights into regularization effects, with experiments on benchmark datasets.
Despite its wide use and empirical successes, the theoretical understanding and study of the behaviour and performance of the variational autoencoder (VAE) have only emerged in the past few years. We contribute to this recent line of work by analysing the VAE's reconstruction ability for unseen test data, leveraging arguments from the PAC-Bayes theory. We provide generalisation bounds on the theoretical reconstruction error, and provide insights on the regularisation effect of VAE objectives. We illustrate our theoretical results with supporting experiments on classical benchmark datasets.