Title:
RAIN and NCS5 Benchmarks

  by Milind Zirpe

 Home  |  Back to Papers
 The primary objective of the Brain Computation lab at University of Nevada, Reno
is to discover principles and develop models of social intelligence in an artif
icial agent, with biology as basis. To help with this aim, a complex and relativ
ely biologically realistic spiking neural network simulator was developed. This
is the NeoCortical Simulator version 5 (NCS 5) which is capable of efficiently s
imulating large neural networks (more than 10,000 cells and 1,000,000 synapses)
using a parallel cluster. The work done in this thesis develops neural network m
odels which exhibit the principle of background activity present in a live biolo
gical brain using NCS. The principle of Recurrent Asynchronous Irregular Network
(RAIN) might provide a basis for developing more advanced human aspects of memo
ry, learning, consciousness, and pattern recognition and various other applicati
on fields. Furthermore, benchmarks were done to test neural networks used in a V
irtual Social Robot (VSR) loop. Results of these benchmarks showed capabilities
of our cluster and current software which would prove vital for future upgrades
and design of neural network models using NCS.
Posted on:28 Dec 2007