Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： The success of AlphaGo has made deep learning methods widely concerned in the field of computer games. As we know, supervised training based on deep learning relies on a high-quality dataset consisting of a large amount of manually calibrated samples. However, many non-popular computer games are facing the problem of lacking human-game records as training samples. Therefore, how to generate a reasonably calibrated dataset of configuration data before using deep learning has significant value. In this paper, a data hashing and de-emphasis, and a configuration calibrated method are proposed for the dots and boxes game. According to the characteristics of configuration data at different stages, the proposed method makes use of full Alpha-Beta search, back-tracing search, parallel MCTS algorithm as well as symmetric flip extension to collect massive configuration data as training dataset. Experiment generates 15 million samples in total as the dataset to drive the supervised training model based on deep learning. In addition, the method proposed in this paper also provides valuable reference for the acquisition of training data of other chess games.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： Chinese word segmentation is a fundamental task in Chinese natural language processing. Currently the mainstream methods for Chinese word segmentation exploit statistical machine learning models. These methods usually require manual-annotated segmented sentences as training corpus, yet have neglected the annotated large-scale lexicon resources which have been built before, where these resources can be highly valuable when cross-domain evaluation is conducted, as the gold-standard sentence-level annotations arerare. Recently, the integration of lexicon formation into word segmentation models has gained increasing interest. As a whole, the integration methods can be classified into two categories: one being based on character-based models that cast word segmentation problem as sequence labeling, and the other being based on word-based models that use beam-search to decode. In this paper, we compare these two models, and combine them. Experimental results on benchmark data sets show that lexicon information can be more fully explored after combination, and finally the combined model can achieve better performances with both in- and cross-domain settings.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： The signal interference problem in this high-density deployment of Wi-Fi like banks, offices, mall, it is the pain point restricts user experience and network quality. Firstly, studied the interference detection technology, and proposing an interference level evaluation model to recognize and measure the interference of Wi-Fi nodes. Then, put forward the strategy of "tolerance & avoidance" and the tolerance policy bases on the capture effect theory and wireless resource management technology. It improved the quality of data communication under interference condition. The collision avoidance strategy adopted the local channel self-coordination algorithm and the decentralized distributed architecture to solve the conflict problem in the homogeneous interference. Finally, it implemented the mechanism WifiAAS. Test results show it can improve the performance of 10% and not too expenses.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-19 Cooperative journals: 《计算机应用研究》
Abstract： Control obfuscation can greatly increase the difficulty on detecting implementation path of Android application, and also is the one of the main difficulties on current code static analysis. This article had carried on the research to this question and designed a system DOCFDroid used to solve this problem. The system obtained the CFG relation matrix in the pretreatment stage, and used the depth-first algorithm to find the set of the path. According to the set of source points and the set of end points, the coarse path was obtained. The algorithm of weight screening could effectively obtain the target path set. Based on the sample set DroidBench1.2, the test sample set was further constructed to verify the validity of this method. The experimental results show that the system can effectively resist the interference caused by control obfuscation, and the recognition rate of target path can reach 95.31%.
Hits
Hits
Downloads
Comment