Information Theoretic Learning: New Perspectives in Signal Separation over Finite Fields and Inversion of Wiener Systems

Abstract

This project deals with two key problems in signal processing, which are addressed according to the unifying paradigm of information theoretic learning. In accordance with a broader perspective that adopts bioinspired metaheuristcs and descriptors that are not necessarily differentiable, we have as general objective the development and extension of methods to be applied in the problems of signal separation over finite fields and inversion of Wiener systems. Both problems have a considerable potential of application to several real-world scenarios, leading to interesting perspectives in areas such as genomics, data mining, and seismic processing.This research theme is a natural sequence to the candidate's doctoral research, which is in the stage of conclusion. Therefore, we intend to significantly extend the already accomplished contributions, both in the sense of deepening the proposed paradigms and of analyzing new theoretical / practical possibilities. A summary of the work performed by the candidate during his doctoral studies is found in [1-9].In summary, it is expected that the contributions strengthen the conception that information theoretic learning algorithms must not necessarily be restricted to continuous signals and gradient based adaptation algorithms.