The clustered causal state algorithm efficient pattern discovery for lossy data-compression applications

Clustered Causal State Algorithm (CCSA), a pattern discovery algorithm, is developed in lossy video compression to approximate E-machines, for use in real time and resource limited applications. CCSA performs unsupervised pattern discovery, producing pattern descriptions with computational efficiency for use in data compression in exchange for a small loss in description fidelity. It is based on the hierarchical agglomerative clustering method and attempts to describe patterns intrinsic to a process, which it achieves at a lower computational cost. The inputs to the CCSA program are the symbol stream and the algorithm executes in the following steps: initialization, clustering, finalization. CCSA has the distinct advantage of polynomial computational complexity, and using this algorithm image compression takes few seconds and it could reliably generate 10 to 20 fold compressions.