VARIABLE SENSITIVITY PHOTODETECTOR FOR OPTICAL NEURAL NETWORKS

A novel type of a photodetector called a variable sensitivity photodetector (VSPD) has been developed for an optical implementation of a neural network. The function of the VSPD is a photodetector having both functions of the spatial light modulator and the photodetector. The VSPD with a metal-semiconductor-metal structure is shown to be very suitable for the implementation of the analog synaptic weight with bipolar value. Two-dimensional 8 x 8 array of VSPD's is fabricated on a GaAs substrate and integrated with an LED array with eight lines in hybrid form. This chip can simulate 8-neuron system with 64 variable synaptic connections. The application to pattern classification is demonstrated by using the chip. The chip is used to classify 12 patterns into three categories by the backpropagation learning algorithm in three layered network. It is studied how the nonlinear and nonuniform characteristics of the VSPD array device affect the learning performance.