Neural Markovian Predictive Compression: An Algorithm for Online Lossless Data Compression

Abstract

This work proposes a novel practical and general-purpose lossless compression algorithm named Neural Markovian Predictive Compression (NMPC), based on a novel combination of Bayesian Neural Networks (BNNs) and Hidden Markov Models (HMM). The result is an interesting combination of properties: Linear processing time, constant memory storage performance and great adaptability to parallelism. Though not limited for such uses, when used for online compression (compressing streaming inputs without the latency of collecting blocks) it often produces superior results compared to other algorithms for this purpose. It is also a natural algorithm to be implemented on parallel platforms such as FPGA chips.