Universal Data Compression with Circular Context Trees

Within the ITUL project, funded by an ERC Consolidator Grant, we have developed a data compression algorithm based on circular context trees. The proposed algorithm is universal in that it can compress any type of data.

The proposed compression algorithm offers unprecedented compression performance, and shows an average of at least 82% improvement over commercial algorithms like Lempel-Ziv (ZIP), Burrows-Wheeler transform compression (BZIP), and more complex state-of-the-art algorithms like prediction by partial matching (PPM) and context tree weighting (CTW).

The proposed algorithm has linear decoding complexity.

The proposed PoC is intended to extend the current binary implementation to arbitrary alphabets, provide fast encoder/decoder implementation, develop specific applications to compress satellite observation data and genomic data, develop an indexed version of the compression algorithm in order to access some of the data without full decompression, and investigate specific software licensing options and opportunities.