Autoencoding With a Classifier System

Autoencoders are data-specific compression algorithms learned automatically from examples. The predominant approach has been to construct single large global models that cover the domain. However, training and evaluating of increasing size comes at price additional time computational cost. Conditional computation, sparsity, model pruning techniques can reduce these costs while maintaining performance. Learning classifier systems (LCS) a framework for adaptively subdividing input spaces into an ensemble simpler local approximations together LCS perform conditional computation through use population individual gating/guarding components, each associated with approximation. This article explores decompose domain collection small autoencoders where solutions different complexity may emerge. In addition benefits in convergence cost, it is shown possible code as well resulting decoder cost when compared equivalent.