Compressing human text.
Episode details
Episode description
I will present a new data compression algorithm that compresses human text more effectively than all current variants of PPM, CTW, DMC, LZ (Lempel-Ziv), LZMA, CSE and BWT. It uses a hierarchical non-parametric sequence model and online gradient optimisation. See more at https://www.microsoft.com/en-us/research/video/compressing-human-text/