There exists a near link between machine learning and compression. A procedure that predicts the posterior probabilities of the sequence given its overall heritage may be used for optimal facts compression (by utilizing arithmetic coding over the output distribution). We're constructing Secur