An abnormity detection method for a Go AI chess manual file in an SGF format

Abstract

The invention relates to an abnormity detection method for a Go AI chess manual file in an SGF format. The method comprises the steps that S1, importing text data of the SGF file; S2, processing the text data on the basis of a word bag model to obtain a series of word sets, and calculating the weight of each word on the basis of the obtained sets; S3, combining the weight sets of the words to obtain language model vectors of the text data; and S4, carrying out anomaly detection by utilizing an anomaly detection algorithm, and outputting a detection result. Compared with the prior art, the method has the advantages that the text is converted into the corresponding vector form based on the word bag model, so that the diagnosis can be carried out by means of some existing abnormal diagnosis algorithms, and the machine diagnosis of the chess manual file can be realized.

Description

technical field [0001] The invention relates to an abnormality detection method, in particular to an abnormality detection method for Go AI game record files in SGF format. Background technique [0002] Artificial Intelligence (AI) has been vigorously developed along with many applications in human reality scenarios, and the progress of artificial intelligence in Go has also achieved great results. SGF is the file format for saving the game record information of Go AI, and it is an important file information for Go AI weight training through the input residual network. For a long time, the number and scale of chess record files in SGF format and the complexity of chess skills have continued to grow, but Go AI cannot eliminate and identify abnormal game records in SGF format files. The abnormal game record in the SGF format file seriously affects the training data of the residual network, which leads to the expansion of the weight scale of the inferior Go weight generated by...

AI Technical Summary

Problems solved by technology

[0005] Among the many Go AIs today, the network training of weights requires more than 100,000 SGF game records, and hundreds of abnormal SGF game records are often ignored. The training data of the residual network is used to generate weights to improve the Go AI's chess performance. The weights generated by unprocessed abnormal game record training will cause Go AI to have many blind spots when playing the game, and its chess strength will be greatly reduced, and the abnormal SGF game record will consume a lot of weight training time and cause superpower consumption of GPU and other hardware

Images