Neuro-genetic based method to the classification of acupuncture needle: a case study

摘要

The paper presents a method which combines neural network (NN) and genetic algorithm (GA) to identify the defective acupuncture needle, which are extensively used in medical areas. Neural network is one of the powerful tools for pattern classification where the only data available are the inputs and outputs. Instead of using the popular backpropagation (BP) algorithm, genetic algorithm is investigated to optimize the weights and biases of neural network to avoid being stuck into local minima. A new neural network specific crossover operator is implemented to cope with the competing conventions problem, where the neural network has functional similar hidden nodes. The benefits of this genetic operator and the effects of other genetic parameters on the search space are analyzed using the real acupuncture needle data gathered from CCD Visual System.