Research On Strong Group Quantization In Computer Go

Computer game is an important area of artificial intelligence research, and it is animportant aspect of the development level of artificial intelligence. Computer Go is abranch of computer game.It is a emphasis that how to quantize the value of strong group. According to theknowledge structure and presentation of Go players, combined with related algorithms,a series of solution of the strong group quantization is proposed. It is not only helpfulfor calculating the value of strong group, but also for enhancing the level of the Goprogram, which has innovative significance and utilitarian value in the research anddevelopment of artificial intelligence.The main contents and achievements are as follows:(1) Two kinds of parameter optimization algorithm based on Fence CognitiveAma3model are proposed. Rise Drop optimization algorithm and Bisectionoptimization algorithm are designed. A traversal method is used for searching theoptimal parameter combination in Rise Drop algorithm. Rise Drop algorithm reducedthe redundancy calculation and improved the operating speed of the algorithm. Dividethe interval method is used to find the optimal parameter combination in Bisectionalgorithm. In accordance with the bound of parameter to divide the interval, thencontinue to divide the region of parameter which the error rate is minimum until theinterval reduced to a set threshold. The results show that two kinds of optimizationalgorithms have reached a high accuracy to quantify the strong group value.(2) Based on A*algorithm, the Thick-Solid quantitative model is designed. Firstly,due to the shortage of strength influence functions that can touch the empty spot thatbehind the other pieces, a dispersion of thought that the influence value need to bypassobstacles is proposed. A*algorithm is used to obtain the optimal path. Secondly, designa influence function named Radiate to calculate the influence value of single stone, andbuild Thick-Solid model. Finally, genetic algorithms is used to optimize the parameters.The experimental result shows the model is of high accuracy so that it can beapplied in starting game, middle game and end game of the computer Go and providethe basis for the static evaluation program.(3) ElegantGo, a strong group quantization testing system is developed by integrating the above. It is not only helpful for testing the performance of Fence RiseDrop model, Fence Bisection model and Thick-Solid model, but also for speeding upthe testing process by avoiding the tedious process of manually input Go files.Simultaneously, the system is user-friendly to visualize the test results and easy todebug.