Data compression involves two separate procedures: source modeling and encoding. Many of the existing compression schemes use domain- specific knowledge to attempt to simplify the process and optimize the compression, and in the process of doing so, they combine modeling and encoding into one procedure, and this rigid structure renders the compression schemes not ready for change. In this paper, we discuss and implement a Model-Free Encoding Compression Systems to provide a flexible compression architecture by using probabilistic graphical modeling. This architecture separates the two concepts of source modeling and encoding by implementing a lightweight encoder and a probabilistic-inferential decoder that applies its knowledge about the source model during the decoding process.

I am a senior studying Mathematics and Computer Science, with a concentration in Artificial Intelligence. In this SuperUROP project, which uses inference on graphical models to solve the problem of data compression, I seek to bring theoretical results of information theory to build a product that solves one of the most fundamental problems of modern computing: compression.