PRACTICAL OPTOELECTRONIC NEURAL-NETWORK REALIZATIONS BASED ON THE FAULT-TOLERANCE OF THE BACKPROPAGATION ALGORITHM

This paper describes how the fault tolerance of the backpropagation al gorithm can be used to accommodate the realistic (nonideal) transfer c haracteristics of the optical communication links used, between neural layers, in optoelectronic neural networks. In particular we demonstra te that networks, utilizing MSM (metal-semiconductor-metal) photodiode s (PD's) and either LED (light emitting diode) or MQW (multiple quantu m well) Plaser transmitters within these intraneural links, are capabl e of performing satisfactorily even in the presence of such nonideal d evice phenomena as: 60% optical crosstalk, 50% optoelectronic device v ariation, or a thresholded (I-th = 0.5 I-max) laser output character istic. Subsequent to this, we then show how it is possible to use this fault tolerance to simplify the neuron architecture, to the extent th at it consists only of MSM PD's, a current amplifier, and an MQW laser . The overall neuron transfer function is then a first-order approxima tion to the original sigmoidal function.