Constructive Learning of Binary Neural Networks and Its Application to Nonlinear Shift Register Synthesis

摘要

In this paper, a constructive learning algorithm is proposed to train a three-layer binary neural network (BNN) for the generation of arbitrary binary-to-binary mapping. Then the algorithm is used to generate the feedback function in nonlinear shift register synthesis. A nonlinear shift register synthesis algorithm is given which can obtain both nonlinear complexity of the given sequence and feedback function of the shift register simultaneously. The algorithm is simple, reliable and it has small burden of storage and computation. The convergence of this algorithm is completely guaranteed. Furthermore, the proposed algorithm and the three-layer BNN have simple implementation in VLSI technology.