Learning Deep Representations of Data Distributions

This open-source textbook bridges classical analytical models like PCA, ICA, and Dictionary Learning with modern deep networks through the unifying principle of data compression.

It covers learning low-dimensional structures from high-dimensional data, utilizing them for Bayesian inference in tasks like image classification, completion, segmentation, and generation.

The book provides theoretical foundations, mathematical verifications, computational methods, and practical applications with code examples for large-scale datasets.