Learned Lossless Compression via an Extension of the Bayes Codes

Yuta Nakahara, Shota Saito, Koshi Shimada, Toshiyasu Matsushima

Keywords: Learned compression, lossless compression, Bayes codes

TL;DR: We introduce a framework to incorporate an idea of learned compression into the Bayes codes.

Abstract: Learned compression is an emerging scheme of data compression where the encoder and decoder are learned from data. In learned compression, various deep neural network-based methods are proposed. However, the idea of data-based code construction does not necessarily require deep neural networks. In this paper, we introduce a framework to incorporate this idea into classical codes via an extension of the Bayes codes.

Submission Number: 5