PPGINF PÓS-GRADUAÇÃO EM ENGENHARIA DA INFORMAÇÃO FUNDAÇÃO UNIVERSIDADE FEDERAL DO ABC Teléfono/Ramal: No informado http://propg.ufabc.edu.br/ppginfo

Banca de QUALIFICAÇÃO: RENNAN SANTOS DE ARAUJO

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
DISCENTE : RENNAN SANTOS DE ARAUJO
DATA : 05/06/2019
HORA: 14:00
LOCAL: sala 304, 3º andar do Bloco B, no campus Santo André
TÍTULO:

Memristive Vanadium Dioxide (VO2) Neuron Model Based on PRBS-PWM Codified Spikes


PÁGINAS: 55
GRANDE ÁREA: Engenharias
ÁREA: Engenharia Elétrica
SUBÁREA: Circuitos Elétricos, Magnéticos e Eletrônicos
ESPECIALIDADE: Circuitos Lineares e Não-Lineares
RESUMO:

Neuromorphic computing represents a paradigm of computational architecture that, based on the complex mechanics of activities of a biological neural system, aims to be a more efficient alterative for the large-scale information processing, "threatening" the prevailing hegemony of the consolidated Architecture of von Neumann. Historically, the reduction of the energy cost of computational operations has always been a factor of great relevance in the development of new technologies and this is, perhaps, the main beneficial factor of neuromorphic computing. In this bioinspired architecture, all information manipulation is carried out through artificial neural networks, which are constituted of "neurons" and numerous interconnections (synapses) between them, which contribute to the data being processed and persisted in a fast, accurate, energy-efficient and resistant-to-local-faults way.

Based on these concepts, this work proposes a new model of neuron, stimulated by spikes codified by PRBS (Pseudorandom Binary Sequence) and modulated by Pulse Width Modulation (Pulse Width Modulation), whose memristence and activation are based on the behavior of the thermal dynamics and non-linear phase transition of the inorganic metal oxide compound Vanadium Dioxide (VO2). It is conjectured that this new neuron paradigm, when implemented in an artificial neural network, will be able to establish a higher energy saving index, synaptic plasticity and noise resistance, making the neural system more robust and connective.

The simulations carried out in this work demonstrated that the implementation and combination of presynaptic logical operators, together with regulated variations of non-correlation between the input sequences, produce a more solid control of the neuron's activation, generating a higher energy efficiency in information processing.


MEMBROS DA BANCA:
Membro Titular - Examinador(a) Interno ao Programa - 2196309 - CARLOS ALBERTO KAMIENSKI
Membro Titular - Examinador(a) Externo ao Programa - 1672981 - FRANCISCO JAVIER ROPERO PELAEZ
Presidente - Interno ao Programa - 1205456 - LUIZ ALBERTO LUZ DE ALMEIDA
Membro Suplente - Examinador(a) Externo ao Programa - 1545354 - RICARDO CANELOI DOS SANTOS
Membro Suplente - Examinador(a) Interno ao Programa - 1761107 - RICARDO SUYAMA
Notícia cadastrada em: 10/05/2019 09:07
SIGAA | UFABC - Núcleo de Tecnologia da Informação - ||||| | Copyright © 2006-2024 - UFRN - sigaa-2.ufabc.int.br.sigaa-2-prod