Space-variant interconnections based on diffractive optical elements for neural networks: architectures and cross-talk reduction.

Optical architectures for fully connected and limited-fan-out space-variant weighted interconnections based on diffractive optical elements for fixed-connection multilayer neural networks are investigated and compared in terms of propagation lengths, system volumes, connection densities, and interconnection cross talk. For a small overall system volume the limited-fan-out architecture can accommodate a much larger number of input and output nodes. However, the interconnection cross talk of the limited-fan-out space-variant architecture is relatively high owing to noise from the diffractive-optical-element reconstructions. Therefore a cross-talk reduction technique based on a modified design procedure for diffractive optical elements is proposed. It rearranges the reconstruction pattern of the diffractive optical elements such that less noise lands on each detector region. This technique is verified by the simulation of one layer of an interconnection system with 128 x 128 input nodes, 128 x 128 output nodes, and weighted connections that fan out from each input node to the nearest 5 x 5 array of output nodes. In addition to a significant cross-talk reduction, this technique can reduce the propagation length and system volume.

[1]  C C Guest,et al.  Iterative encoding of high-efficiency holograms for generation of spot arrays. , 1989, Optics letters.

[2]  Ashok V. Krishnamoorthy,et al.  A prototype 3D optically interconnected neural network , 1994 .

[3]  David C. Van Essen,et al.  Information processing strategies and pathways in the primate retina and visual cortex , 1990 .

[4]  B K Jenkins,et al.  Subtracting incoherent optical neuron model: analysis, experiment, and applications. , 1990, Applied optics.

[5]  A F Gmitro,et al.  Computer-generated holograms for optical neural networks: on-axis versus off-axis geometry. , 1993, Applied optics.

[6]  M R Taghizadeh,et al.  Binary surface-relief gratings for array illumination in digital optics. , 1992, Applied optics.

[7]  Frank Wyrowski,et al.  Diffractive optical elements: iterative calculation of quantized, blazed phase structures , 1990 .

[8]  B K Jenkins,et al.  Architectural implications of a digital optical processor. , 1984, Applied optics.

[9]  S. Esener,et al.  A scalable optoelectronic neural system using free-space optical interconnects , 1992, IEEE Trans. Neural Networks.

[10]  M. R. Feldman.  Diffractive optics move into the commercial arena , 1994 .

[11]  W. B. Veldkamp.  Wireless focal planes 'On the road to amacronic sensors' , 1993 .

[12]  H. J. Caulfield,et al.  Optical neural networks , 1989 .

[13]  B. K. Jenkins,et al.  Photonic Implementations of Neural Networks , 1992, Optical Computing.

[14]  A F Gmitro,et al.  Design and analysis of fixed planar holographic interconnects for optical neural networks. , 1992, Applied optics.

[15]  A A Sawchuk,et al.  Design of diffractive optical elements with optimization of the signal-to-noise ratio and without a dummy area. , 1997, Applied optics.

[16]  B K Jenkins,et al.  Visual cortex operations and their implementation using the incoherent optical neuron model. , 1993, Applied optics.

[17]  M P Dames,et al.  Efficient optical elements to generate intensity weighted spot arrays: design and fabrication. , 1991, Applied optics.

[18]  R. Gerchberg.  A practical algorithm for the determination of phase from image and diffraction plane pictures , 1972 .