Self-learning estimation method based on RBF neural network in Einstein chess

Abstract

The invention discloses a self-learning estimation method based on the RBF neural network in Einstein chess. According to the self-learning estimation method, the RBF neural network is applied to theestimation method, by means of an MRAN learning rule of the RBF neural network, clustering is carried out on the features of the chessboard, parameters are gradually adjusted, and the board surface estimation of the chessboard is obtained. According to the method, the RBF neural network is used as an estimation function of the Einstein chess, so that the Einstein chess system has on-line learningcapability, the mode and weight parameters in the Einstein chess can be independently learned, thus the problems that the knowledge representation of the estimation function of the existing Einstein chess is not comprehensive and the manual parameter adjustment is complicated are solved, and the precision of the estimation function is improved.

Description

technical field [0001] The invention belongs to the research field of machine games of board games, in particular to a self-learning evaluation method based on RBF neural network in Einstein chess. Background technique [0002] The research on Artificial Neural Network (Artificial Neural Network) started as early as the middle of the twentieth century. People got inspiration from the process of biological neuron transmission and processing information, and hoped that computers could imitate human intelligence. After several ups and downs, the Perception model is the most representative neural network structure. [0003] Radial Basis Function (RBF for short) has good applications in the solution of multivariate difference problems, the numerical solution of partial differential equations, and the construction of neural networks. In 1988, the Radial Basis Function Neural Network (RBFNN) was proposed, and the RBF neural network has a good function fitting effect. Later, some ...

AI Technical Summary

Problems solved by technology

This method can achieve better results in most cases, but due to the variability of the Einstein chess situation and the influence of uncertain factors, the evaluation function cannot make correct judgments in some special cases, knowledge representation Incomplete problems still exist

An incompletely designed valuation function will bias the search in the wrong direction in some cases, leading to a decline in the level of the game

Images