Information and Entropy in Neural Networks and Interacting Systems

摘要: In this dissertation we present a study of certain characteristics interacting systems that are related to information. The first is periodicity, correlation and other information-related properties neural networks integrate-andfire type. We also form quasiclassical quantum generalizations such identify the similarities differences with classical prototype. indicate why entropy may be an important concept for network generalization definition required. Like networks, large ensembles similar units interact need information-theoretic concepts. extend Shannon in novel way, which relevant when have systems, show how it differs from generalizations, as Tsallis entropy. stochasticity arise interactions entangled environment system terms Shannon’s generalized entropies differences. Such indicated use prior probability distributions fit data per Bayesian rules. suggest possible versions pattern recognition, principal goal information processing most networks.