Combining Stochastic Competitive Scheme and Hysteresis Quantized Neuron for Reliability Maximization with Budget and Weight Constraints
Jiahai Wang, Yalan Zhou · 2006
In this paper, we propose a new neural network method combining stochastic competitive scheme and hysteresis quantized neurons for the reliability optimization of a series system with multiple-choice constraints incorporated at each subsystem, to maximize the system reliability subject to the system budget and weight. In the proposed algorithm, the neurons are divided into two classes: One is binary neurons with stochastic competitive scheme and the other is quantized neurons with hysteresis. The competitive scheme always provides a feasible solution and search space is greatly reduced without a burden on the parameter tuning. Furthermore, the stochastic dynamics and hysteresis can help the neural network escape from local minima, and therefore the proposed algorithm can get better results than other neural network method.