|
| |
|
Neuro-fuzzy soft computing refers to massively
interconnected, parallel computing approaches inspired by biological
systems. Soft computing is a partnership of methodologies that play
pivotal roles in the conception, design and application of intelligent
systems.
Idaho's research goal is to address theoretical and practical
issues for developing cost-effective systems by combining digital and
analog technology in a mixed-signal design. This six-member team from
UI (both Moscow and Boise campuses) and ISU includes researchers
experienced in VLSI design and implementation of neuro-fuzzy systems,
adaptive logic system design, and genetic and soft-computing
algorithms, taking multiple approaches to this challenge and
investigating an exciting new paradigm largely unexplored in neural
networks.
|
|
Bo
Liu, UI graduate student
|
|
|
|
Faculty members names link directly to their websites.
[Back To Top]
|
Published |
|
A CMOS Neuron for VLSI Circuit
Implementation of Pulsed Neural Networks, Liu, B., J.F. Frenzel,
The 28th Annual Conference of the IEEE Industrial Electronics
Society (IECON'02), pp.3182-3185, 2002. |
|
Behavioral diversity and a probabilistically optimal GP ensemble,
Imamura, K., T. Soule, R.B. Heckendorn, and J.A. Foster, Genetic
Programming and Evolvable Machines, vol. 4, pp.235-254, 2003. |
|
A versatile pulse-mode biomimic
artificial neuron using a capacitor-free integrate-and-fire technique,
Barnes, B.C., R.B. Wells, Proc. 29th Ann. Conf. Ind. Elec.
(IECON'03), vol. Nov. 2-7, 2003. |
|
Non-Linearly separable Cluster
Classification: An Application for a Pulse-Coded CMOS Neuron,
Konduri, S., and J.F. Frenzel, ANNIE'03, 2003. |
|
Capacitor-Free Leaky Integrator for
Biomimic Artificial Neurons, Wells, R.B. and B.C. Barnes, IEE
Electronics Letters, vol. 38, no. 17, pp. 974-976, 2002. |
|
Glide Algorithm with Tunneling: A fast, reliably-convergent
algorithm for neural network training, Kantabutra, V., B.
Tsendjav+, and E. Zheleva+, ANNIE, pp. TP2.2E, 2003. |
|
Evolving a Strongly Recurrent
Neural Network to Simulate Biological Neurons, Soule, T., Y. Chen,
and R.B. Wells, Proc. 28th Ann. Conf. IEEE Indus. Elec. Soc,
Nov. 5-8, 2001, pp. 3191-3195, 2002. |
|
Co-evolving faults to improve the fault tolerance of sorting
networks, Harrison, M., and J.A. Foster, Proc. European
conference on genetic programming, vol. 3003, pp. 57-66, 2004. |
|
Improving the Survivability of a Simple Evolved Circuit through
Co-evolution, Harrison, M., and J.A. Foster, NASA/IEEE
Conference on Evolvable Hardware, vol. 6/24/04, pp. 123-129,
2004. |
|
Non-Linearly separable Cluster Classification: An Application for a
Pulse-Coded CMOS Neuron, Konduri, S., and J.F. Frenzel, ANNIE
2003, vol. 13, pp. 63-67, 2003. |
|
Accepted |
|
Efficient, Practical Adders for FPGAs, Kantabutra, V., S. Perri,
M. Iachino, and P. Corsonello, Circuit Cellar Magazine, 2002. |
+ Undergraduate
authors
[Back To Top]
|
EPSCoR
Investigators |
Agency
Title |
Dates |
Amount |
|
J.A.
Foster
P. Joyce |
National Science Foundation
Undergraduate scholarships for computer science and mathematics |
2003-08 |
$396,000 |
|
J.A.
Foster |
National Science Foundation-
ITR
Tribal Law Enhancement Project |
2003-08 |
$410,000 |
|
R.B. Wells |
Hewlett Packard Co.
Laser Electrophotography
Research Program |
2002-03 |
$199,724 |
|
R.B. Wells
J.F. Frenzel
T. Soule
V. Kantabutra |
National Science Foundation
REU Site: Computational Neuroscience & Technology
Research Experience for Undergraduates |
2003-05 |
$232,222 |
|
R.B.
Wells
D. McIlroy
W.J. Yeh |
Office of Naval Research
Advanced Microwave Ferrite Research: Phase One |
2004-2005 |
$1,250,000 |
[Back To Top]
-
Development of a simple
evolutionary algorithm for the evolution of weights, within a
recurrent neural networks with complex behavior
-
Development of a parallel
algorithm for the evolution of weights and topologies of recurrent
neural networks
-
Invention of new
capacitor-free all-MOSFET leaky integrator circuit (patent pending)
-
Design of a basic
frequency-modulating biomimic neuron
-
Deduction of three basic
mathematical theorems to be used in implementing biomimic modulation
by second-messenger effects
-
Development of a
neural-network-based computer-aided design tool to assist engineers
involved in the design of formatters for print engines
-
Solved, for one network at
least, the old "flat-region" problem in neural networks
training
[Back To Top]
-
"NSF EPSCoR support
has provided a badly needed "kick-start" to our research in
neurocomputing. As a direct consequence of this support we have been
able to come up with innovative new approaches to neural network
hardware design that appears to offer solutions to a number of size,
speed, and power consumption problems that have balked the development
of significant neurocomputer development."
-
"My approach prior to
this grant had been to work alone and attack some of the most
difficult (but still solvable) or centrally important research
problems in a few fields that interested me... However, by working in
a research group I have been able to learn about neural networks from
colleagues with a lot of experience in the field."
-
"The NSF EPSCoR
support has allowed the training and development of several students
who are now conducting high quality research. This has expanded my
primary research efforts, made it possible to conduct more speculative
research and allowed additional time for further grant writing."
-
"The NSF EPSCoR
support has also made additional travel possible. In particular, being
able to attend conferences beyond those devoted to evolutionary
computation has greatly expanded the potential applications of my
research.
-
"The most significant
accomplishment has been the training of students. Under this project I
have assisted one undergraduate, two masters students, and one Ph.D.
student in their professional development. I expect all three to
continue in a research related career. Due in large part to the
assistance of this grant."
[Back To Top] |
|
