Design and analysis of fixed planar holographic interconnects for optical neural networks

Abstract

A neural-network architecture of multifaceted planar interconnection holograms and optoelectronic neurons is analyzed. Various computer-generated hologram techniques are analyzed and tested for their ability to produce an interconnection hologram with high-accuracy interconnects and high diffraction efficiency. A new technique is developed by using the Gerchberg-Saxton algorithm, followed by a random-search error minimization that produces the highest interconnect accuracy and the highest diffraction efficiency of the techniques tested. Analysis of the system shows that the hologram has the capacity to connect 5000 neuron outputs to 5000 neuron inputs with bipolar synapses and that the encoded synaptic weights have an accuracy of ~5 bits. A simple feedback system is constructed and demonstrated.

Optical implementation of artificial synapses by means of the Wigner distribution approach

“…On the other hand, it can also be defined (7) from the Fourier transform F(w) off(r) by WF(w,r)= fF(w+)F*(w_) exp(ikrw') dw' . (8) (2) where p(r0) denotes the phase function of the DOE. EquaIt is known that, in general, the signal transformation in a tions (7) and (8) …”

<title>Diffractive optical interconnect element design for digital switching</title>

“…Previous authors [3][4][5][6] have documented their results in increasing the SBWP and modifying the Gerchberg I Saxton. Under the current program, reduction in harmonic aliasing is demonstrated by a simple algorithm.…”

“…interconnects have been added, with linear spacing (8 pixels between elements). These 5 new interconnects occupy locations (3,6), (4,6), (5,6), (8,6) and (8,7) as shown in Figure 5a. This is the interconnect constraint applied after the first Fourier transform performed on the modified Gerchberg I Saxton algorithm.…”

Computer-generated holograms for optical neural networks: on-axis versus off-axis geometry

“…Previously we discussed the capabilities and limitations of this optical neural-network system based, which is based on computer-generated holograms. 3 Here we focus on the diffraction geometry. Three diffraction geometries are considered: the standard off-axis geometry, the on-axis geometry, and a modified on-axis geometry.…”