Realization of Neurons with Multiple Thresholds for XOR Operations in BP Networks Based on Fuzzy Genetic Algorithms
Download as PDF
DOI: 10.23977/icmmct.2019.62050
Corresponding Author
Qi Anzhi
ABSTRACT
The principle of realizing multi-threshold neuron nonlinearity by using fuzzy genetic algorithm BP network is analyzed, and the XOR operation by using neurons is taken as an example. To solve the unsolvable problem caused by inappropriate selection of initial values of connection weights and thresholds of BP neural network, a fuzzy genetic algorithm for approximation of global optimal solution and an exact value of local optimal solution are combined. At the same time, the initial weights and thresholds of neurons are optimized by using the fuzzy genetic algorithm, and three basic operations in logic are realized by using a multi-threshold neuron. The multi-threshold neuron of this basic operation can form a multi-threshold neural network that implements arbitrary ternary functions. Since the ability of single neuron information processing is improved, the optimization result is used as the initial value of the BP network and then trained by BP network. The network, so alternately runs the BP network and the fuzzy genetic algorithm until the accuracy required by the problem is reached.
KEYWORDS
Fuzzy genetic algorithm, BP network, Multi-threshold neuron