Postscript viewgraph of Modeling Photorefractive Learning
Cool Images from Tim Slagle's Optical Competitive Learning project
Charles Garvin's research summary
Undergraduate summer research opportunity
We are attempting to apply the organizational principles of neural networks in the brain to artificially constructed adaptive systems made from optical devices. The neurons are implemented using custom integrated circuits that incorporate photodetectors and light modulators, while the adaptive synapses are formed as dynamic holographic interconnection gratings whose strength grows in proportion to the correlated activity of the source and destination neurons. We have developed architectures, devices, and simulations for self-aligning multilayer holographic optical learning systems for the implementation of optical back propagation and optical competitive learning, which are prototypical supervised and unsupervised learning algorithms. These systems are the first self-aligning holographic implementations of multilayer neural networks. In addition we have demonstrated the first winner-take-all VLSI/liquid crystal modulator array with 576 neurons grouped into 31 competitive patches on the appropriate sparse grid topology required for volume holographic learning.
[Journal Publications: 5,
7,
13,
16,
Conference Publications:
11,
12,
13,
14,
15,
18,
28,
33
47
]
Postscript viewgraph of Modeling Photorefractive Learning
Cool Images from Tim Slagle's
Optical Competitive Learning project
Charles Garvin's research summary
Undergraduate summer research opportunity
Return to KAOS Research Page