An Introduction to Neural Data Compression
It provides an introductory review for a broader machine learning audience, covering background in information theory and computer vision, but is incremental as it synthesizes existing literature without presenting new results.
This paper introduces neural data compression as a field where neural networks and machine learning methods are applied to data compression, leveraging recent advances in generative models like normalizing flows and VAEs to learn compression algorithms end-to-end from data.
Neural compression is the application of neural networks and other machine learning methods to data compression. Recent advances in statistical machine learning have opened up new possibilities for data compression, allowing compression algorithms to be learned end-to-end from data using powerful generative models such as normalizing flows, variational autoencoders, diffusion probabilistic models, and generative adversarial networks. The present article aims to introduce this field of research to a broader machine learning audience by reviewing the necessary background in information theory (e.g., entropy coding, rate-distortion theory) and computer vision (e.g., image quality assessment, perceptual metrics), and providing a curated guide through the essential ideas and methods in the literature thus far.