Título: Em busca de estabilização na GasNet
Autores: Santos, Carmen L. R.; Oliveira, Pedro P. B. de; Souza, Celso R.
Resumo: The GasNet is a novel kind of artificial neural network which, in addition to the traditional electric type, point-to-point communication between units, it also allows for interaction between the units through a diffusable chemical modulator. The GasNet is a dynamical system with a strong bias on recurrency, due to the recurrent electric connections it may have between the units, and also because of the coupled electrochemical activity inherent to it. As a consequence, a question arises of the influence of the latter, over GasNet ́s performance and stability. Initial steps towards addressing this question are taken herein, from an empirical standpoint, based on a task carried out by a simulated robot. This experiment, together with our earlier results in the XOR problem, indicate difficulties for GasNet, and provide clues for the identification of its causes, and the measures to correct, or, at least, alleviate it.
Código DOI: 10.21528/CBRN2001-080
Artigo em pdf: 5cbrn_080.pdf
Arquivo BibTex: 5cbrn_080.bib