Author : Associate Prof. Takashi Kohno
Associate Prof. Takashi Kohno
Institute of industrial science
University of Tokyo, Japan
Associate Prof. Takashi Kohno
Institute of industrial science
University of Tokyo, Japan
"Construction of silicon neural network circuit and it's application"
ABSTRACT
Neuro-mimetic circuits are attracting increasing attentions not only as
an imperative tool for bio-silico hybrid systems but also as a candidate
for the basic technology for the next-generation information processing
systems which are expected to be intelligent, autonomous, adaptive, and
power-efficient.
We developed qualitative-modeling based silicon neuronal circuits. Our
quatlitative-modeling approach intends to realize simple and low-power
silicon neuronal circuits that reproduce the overall dynamical properties
of the neuronal activities including spike generation mechanisms that are
ignored in the integrate-and-fire-based models (e.g. Izhikevich model).
In this talk, we introduce our low-power analog silicon neuron circuit,
digital silicon neuron models, and digital silicon neuronal networks (SNN)
constructed on an FPGA. The performance of an auto-associative memory task
executed on our digital SNN was dependent on the spike generation mechanism
of the silicon neurons. It indicates the possiblity that the dynamics of
spike generation plays an important role in the information processing in
the brain.
BIOGRAPHY
He received the B.E. degree in medicine and the Ph.D. degree in mathematical
engineering from University of Tokyo, Japan, in 1996 and 2002, respectively.
He worked on medical care information systems in Hamamatsu University School
of Medicine for two years. Then he took the job of a group leader in Aihara
Complexity Modeling Project, Exploratory Research for Advanced Technology
(ERATO), Japan Science and Technology Agency (JST), Japan. He is at his
current position from 2006.
His research interests include the modeling of spiking neuronal networks
based on the nonlinear dynamics and silicon neural networks. His final goal
is to realize a brain-like computing system that is at least comparable to
the human brain.
Neuro-mimetic circuits are attracting increasing attentions not only as
an imperative tool for bio-silico hybrid systems but also as a candidate
for the basic technology for the next-generation information processing
systems which are expected to be intelligent, autonomous, adaptive, and
power-efficient.
We developed qualitative-modeling based silicon neuronal circuits. Our
quatlitative-modeling approach intends to realize simple and low-power
silicon neuronal circuits that reproduce the overall dynamical properties
of the neuronal activities including spike generation mechanisms that are
ignored in the integrate-and-fire-based models (e.g. Izhikevich model).
In this talk, we introduce our low-power analog silicon neuron circuit,
digital silicon neuron models, and digital silicon neuronal networks (SNN)
constructed on an FPGA. The performance of an auto-associative memory task
executed on our digital SNN was dependent on the spike generation mechanism
of the silicon neurons. It indicates the possiblity that the dynamics of
spike generation plays an important role in the information processing in
the brain.
BIOGRAPHY
He received the B.E. degree in medicine and the Ph.D. degree in mathematical
engineering from University of Tokyo, Japan, in 1996 and 2002, respectively.
He worked on medical care information systems in Hamamatsu University School
of Medicine for two years. Then he took the job of a group leader in Aihara
Complexity Modeling Project, Exploratory Research for Advanced Technology
(ERATO), Japan Science and Technology Agency (JST), Japan. He is at his
current position from 2006.
His research interests include the modeling of spiking neuronal networks
based on the nonlinear dynamics and silicon neural networks. His final goal
is to realize a brain-like computing system that is at least comparable to
the human brain.