Fully Coupled and Feedforward Neural Networks with Complex-valued Neurons

old_uid2794
titleFully Coupled and Feedforward Neural Networks with Complex-valued Neurons
start_date2007/05/10
schedule10h
onlineno
location_infosalle 549
summaryThis talk introduces a definition of complex-valued neurons with discrete outputs. It covers a novel method of their applications in fully coupled associative memories. Such memories are able to process multiple gray levels when applied for image de-noising. In addition, complex-valued neurons can be generalized to take a continuum of values. Learning of such neurons is demonstrated and described in the context of traditional multilayer feedforward network learning. Such learning is derivative-free and it usually requires reduced network architecture. Selected examples and applications of such networks are also discussed.
responsiblesBouchon-Meunier, Diaz, Gallinari