Understanding Variational Autoencoders with Intrinsic Dimension and Information Imbalance
This work provides tools for architecture search and diagnosing underfitting in VAEs, contributing incrementally to understanding deep generative models through geometric analysis.
The paper analyzes Variational Autoencoders (VAEs) using Intrinsic Dimension and Information Imbalance, revealing a behavioral transition when bottleneck size exceeds data ID, characterized by a double hunchback ID profile and distinct training phases of rapid fit and slower generalization.
This work presents an analysis of the hidden representations of Variational Autoencoders (VAEs) using the Intrinsic Dimension (ID) and the Information Imbalance (II). We show that VAEs undergo a transition in behaviour once the bottleneck size is larger than the ID of the data, manifesting in a double hunchback ID profile and a qualitative shift in information processing as captured by the II. Our results also highlight two distinct training phases for architectures with sufficiently large bottleneck sizes, consisting of a rapid fit and a slower generalisation, as assessed by a differentiated behaviour of ID, II, and KL loss. These insights demonstrate that II and ID could be valuable tools for aiding architecture search, for diagnosing underfitting in VAEs, and, more broadly, they contribute to advancing a unified understanding of deep generative models through geometric analysis.