Title: Towards a Single Chip Model of Cortical Cells

Speaker: Paul Hasler, Georgia Institute of Technology

When/where: Monday, March 22nd, 4:00PM, Room 400 AKW

Abstract:
The computational complexity even in a small region of primate cortex is
immense compared to our existing computer technology with power
consumption much less than most handheld electronics. The question that
we will address is looking at building analog IC models that make a
close match between silicon physics and biological physics that will
enable building electrical models of small regions of cortex on a single
programmable and reconfigurable IC. We expect this research to impact
the fundamental understanding of the computation of cortex, and to yield
valuable insight into building adaptive synthetic neural systems.

We will begin by looking at the electrical properties of biological
neurons, and showing that the physical principles governing ion flow in
biological neurons share some interesting similarities to electron flow
through the channels of a MOSFET transistor. We demonstrate this result
through a 6-transistor circuit model (roughly the size of a large logic
gate) that models the channel behaviors in Hodgkin and Huxley's original
neuron
preparation.

Next, we will look at the development of dendritic computation on an
integrated circuit. Utilizing the same transistor modeling of
biological channels, we can build models of dendrites and the synapses
connected to these dendrite elements. This approach is enabled by the
use of our programmable analog circuit technology, which allows the
investigation the learning behavior in realistic dendrites by including
learning synapses into
these dendritic models. Further, these dendrite circuits, when
configured into a two-dimensional array, can be programmed and
configured for studying dendrites with an arbitrary arboration pattern.

As a result of this research, we believe there is a common framework
underlying the computation performed by Hidden Markov Models (HMM) and
the computation performed by cortical dendrites. This approach
illustrates the design synergy of Neuromorphic engineering: biological
systems inspire engineering design, and engineering practice inspires
biological theory. This research provides a good example of how an
integrated-circuit approach can bridge signal processing techniques and
neural modeling; only simultaneous construction of these two different
ICs made similarities obvious.

Bio:
Paul Hasler is an Associate Professor in the School of Electrical and
Computer Engineering at Georgia Institute of Technology. Dr. Hasler
received his M.S. and B.S.E. in Electrical Engineering from Arizona
State University in 1991, and received his Ph.D. from California
Institute of Technology in Computation and Neural Systems in 1997. His
current research interests include low power electronics, mixed-signal
system ICs,
floating-gate MOS transistors, adaptive information processing systems,
"smart" interfaces for sensors, cooperative analog-digital signal
processing, device physics related to submicron devices or floating-gate
devices, and analog VLSI models of on-chip learning and sensory
processing in neurobiology.