We consider the implementation of high capacity Ho-Kashyap associative processors on non- ideal optical and analog VLSI systems. Processor non-idealities considered include quantization, non-uniform beam illumination, and nonlinear device characteristics. New training-out techniques to overcome these non-idealities are advanced. We obtain optimal performance in the presence of stochastic noise by proper selection of the processor parameter (sigma) syn. We derive important results that allow us to a priori determine the optimal value of (sigma) syn and the expected recall accuracy P'c without having to simulate the specific processor. We present a new algorithm that allow us to achieve storage near the theoretical maximum capacity (2N, where N is the dimensionality of the input vector) with excellent recall accuracy. Optical laboratory results are included. We achieved storage of 1.5 N with recall accuracy P'c >= 95% with input noise of standard deviation (omega) 1 equals 0.02 present and with optical analog components with 5 bit input accuracy and 8 bit memory matrix accuracy. With higher accuracy analog VLSI components (10 bit input accuracy and 11 bit weight accuracy), we achieve storage of 1.75 N with P'c equals 96.43%.
[1]
J. Ohta,et al.
GaAs/AlGaAs optical synaptic interconnection device for neural networks.
,
1989,
Optics letters.
[2]
David Casasent,et al.
Ho-Kashyap advanced pattern-recognition heteroassociative processors
,
1990,
Optics & Photonics.
[3]
Tsunehiro Aibara,et al.
Least squares associative memory and a theoretical comparison of its performance
,
1989,
IEEE Trans. Syst. Man Cybern..
[4]
David Casasent.
Multifunctional hybrid optical/digital neural net
,
1990,
Defense, Security, and Sensing.
[5]
D P Casasent,et al.
Ho-Kashyap optical associative processors.
,
1990,
Applied optics.
[6]
David Casasent,et al.
Ho-Kashyap CAAP 1:1 associative processors
,
1991,
Other Conferences.
[7]
Edward A. Rietman,et al.
Back-propagation learning and nonidealities in analog neural network hardware
,
1991,
IEEE Trans. Neural Networks.
[8]
Anilkumar P. Thakoor,et al.
Analog parallel processor hardware for high-speed pattern recognition
,
1990,
Other Conferences.