As shown in AlphaGo, AlphaGo Zero, and AlphaZero, reinforcement learning is effective in learning of evaluation functions (or value networks) in Go, Chess and Shogi. In their training, two procedures are repeated in parallel; self-play with a current evaluation function and improvement of the evaluation function by using game records yielded by recent self-play. Although AlphaGo, AlphaGo Zero, and AlphaZero have achieved super human performance, the method requires enormous computation resources. To alleviate the problem, this paper proposes to incorporate a checkmate solver in self-play. We show that this small enhancement dramatically improves the efficiency of our experiments in Minishogi, via the quality of game records in self-play. It should be noted that our method is still free from human knowledge about a target domain, though the implementation of checkmate solvers is domain dependent.
{"title":"Learning of Evaluation Functions via Self-Play Enhanced by Checkmate Search","authors":"T. Nakayashiki, Tomoyuki Kaneko","doi":"10.1109/TAAI.2018.00036","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00036","url":null,"abstract":"As shown in AlphaGo, AlphaGo Zero, and AlphaZero, reinforcement learning is effective in learning of evaluation functions (or value networks) in Go, Chess and Shogi. In their training, two procedures are repeated in parallel; self-play with a current evaluation function and improvement of the evaluation function by using game records yielded by recent self-play. Although AlphaGo, AlphaGo Zero, and AlphaZero have achieved super human performance, the method requires enormous computation resources. To alleviate the problem, this paper proposes to incorporate a checkmate solver in self-play. We show that this small enhancement dramatically improves the efficiency of our experiments in Minishogi, via the quality of game records in self-play. It should be noted that our method is still free from human knowledge about a target domain, though the implementation of checkmate solvers is domain dependent.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123165547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Evaluation functions are crucial for building strong computer players in two-player games, such as chess, Go, and shogi. Although a linear combination of a large number of features has been popular representation of an evaluation function in shogi, deep neural networks (DNNs) are recently considered to be more promising by the success of AlphaZero in multiple domains, chess, Go, and shogi. This paper shows that three loss functions, loss in comparison training, temporal difference (TD) errors and cross entropy loss in win prediction, are effective for the training of evaluation functions in shogi, presented in deep neural networks. For the training of DNNs in AlphaZero, the main loss function only consists of win prediction, though it is augmented with move prediction for regularization. On the other hand, for training in traditional shogi programs, various losses including loss in comparison training, TD errors, and cross entropy loss in win prediction, have contributed to yield accurate evaluation functions which are the linear combination of a large number of features. Therefore, it is promising to combine these loss functions and to apply them to the training of modern DNNs. In our experiments, we show that training with combinations of loss functions improved the accuracy of evaluation functions represented by DNNs. The performance of trained evaluation functions is tested through top-1 accuracy, 1-1 accuracy, and self-play.
{"title":"Comparison of Loss Functions for Training of Deep Neural Networks in Shogi","authors":"Hanhua Zhu, Tomoyuki Kaneko","doi":"10.1109/TAAI.2018.00014","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00014","url":null,"abstract":"Evaluation functions are crucial for building strong computer players in two-player games, such as chess, Go, and shogi. Although a linear combination of a large number of features has been popular representation of an evaluation function in shogi, deep neural networks (DNNs) are recently considered to be more promising by the success of AlphaZero in multiple domains, chess, Go, and shogi. This paper shows that three loss functions, loss in comparison training, temporal difference (TD) errors and cross entropy loss in win prediction, are effective for the training of evaluation functions in shogi, presented in deep neural networks. For the training of DNNs in AlphaZero, the main loss function only consists of win prediction, though it is augmented with move prediction for regularization. On the other hand, for training in traditional shogi programs, various losses including loss in comparison training, TD errors, and cross entropy loss in win prediction, have contributed to yield accurate evaluation functions which are the linear combination of a large number of features. Therefore, it is promising to combine these loss functions and to apply them to the training of modern DNNs. In our experiments, we show that training with combinations of loss functions improved the accuracy of evaluation functions represented by DNNs. The performance of trained evaluation functions is tested through top-1 accuracy, 1-1 accuracy, and self-play.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"122 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133830723","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yang Liu, Guijuan Zhang, Xiaoning Jin, Yaozong Jia
The personalized video recommendation system provides users with great convenience while surfing in the video websites. Among many algorithms adopted by recommendation system, the collaborative filtering algorithm is the most widely used and has achieved great success in practical applications, however, the recommended performance suffers from the problem of data sparsity severely. We propose a model that adopts Doc2Vec to deal with video's text information and integrates genre information into rating matrix pre-padding to reduce the sparsity of ratings. The experimental results show that pre-padding ratings is of high quality and the algorithms based on collaborative filtering achieve better performance on the padded datasets.
{"title":"Rating Matrix Pre-Padding for Video Recommendation","authors":"Yang Liu, Guijuan Zhang, Xiaoning Jin, Yaozong Jia","doi":"10.1109/TAAI.2018.00044","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00044","url":null,"abstract":"The personalized video recommendation system provides users with great convenience while surfing in the video websites. Among many algorithms adopted by recommendation system, the collaborative filtering algorithm is the most widely used and has achieved great success in practical applications, however, the recommended performance suffers from the problem of data sparsity severely. We propose a model that adopts Doc2Vec to deal with video's text information and integrates genre information into rating matrix pre-padding to reduce the sparsity of ratings. The experimental results show that pre-padding ratings is of high quality and the algorithms based on collaborative filtering achieve better performance on the padded datasets.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"386 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116125211","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Satoshi Yamada, D. Kitakoshi, Akihiro Yamashita, Kentarou Suzuki, Masato Suzuki
This study aimed to examine the application of an Intelligent Dialogue Agent (IDA) in preventive care frameworks for healthy older adults. Introducing the agent increases familiarity with the frameworks, encourages performance of preventive care exercises, and helps older adults turn using the frameworks into a habit. We used a questionnaire to collect data on older adults' impressions of Information Technology (IT) devices, smart speakers in particular (main components of the IDA), and interviewed the participants after they actually used the smart speaker in order to discuss required functions and expected roles in developing the IDA. Results from the questionnaire and interview revealed promising characteristics of smart speakers and problems concerning Japanese speech recognition.
{"title":"Development of an Intelligent Dialogue Agent with Smart Devices for Older Adults: A Preliminary Study","authors":"Satoshi Yamada, D. Kitakoshi, Akihiro Yamashita, Kentarou Suzuki, Masato Suzuki","doi":"10.1109/TAAI.2018.00020","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00020","url":null,"abstract":"This study aimed to examine the application of an Intelligent Dialogue Agent (IDA) in preventive care frameworks for healthy older adults. Introducing the agent increases familiarity with the frameworks, encourages performance of preventive care exercises, and helps older adults turn using the frameworks into a habit. We used a questionnaire to collect data on older adults' impressions of Information Technology (IT) devices, smart speakers in particular (main components of the IDA), and interviewed the participants after they actually used the smart speaker in order to discuss required functions and expected roles in developing the IDA. Results from the questionnaire and interview revealed promising characteristics of smart speakers and problems concerning Japanese speech recognition.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127652554","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The machine reading comprehension problem aims to extract crucial information from the given document to answer the relevant questions. Although many methods regarding the problem have been proposed, the similarity distraction problem inside remains unsolved. The similarity distraction problem addresses the error caused by some sentences being very similar to the question but not containing the answer. Named entities have the uniqueness which can be utilized to distinguish similar sentences to prevent models from being distracted. In this paper, named entity filters (NE filters) are proposed. NE filters can utilize the information of named entities to alleviate the similarity distraction problem. Experiment results in this paper show that the NE filter can enhance the robustness of the used model. The baseline model increases 5% to 10% F1 score on two adversarial SQuAD datasets without decreasing the F1 score on the original SQuAD dataset. Besides, by adding the NE filter, other existing models increases 5% F1 score on the adversarial datasets with less than 1% loss on the original one.
{"title":"Named Entity Filters for Robust Machine Reading Comprehension","authors":"Yuxing Peng, Jane Yung-jen Hsu","doi":"10.1109/TAAI.2018.00048","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00048","url":null,"abstract":"The machine reading comprehension problem aims to extract crucial information from the given document to answer the relevant questions. Although many methods regarding the problem have been proposed, the similarity distraction problem inside remains unsolved. The similarity distraction problem addresses the error caused by some sentences being very similar to the question but not containing the answer. Named entities have the uniqueness which can be utilized to distinguish similar sentences to prevent models from being distracted. In this paper, named entity filters (NE filters) are proposed. NE filters can utilize the information of named entities to alleviate the similarity distraction problem. Experiment results in this paper show that the NE filter can enhance the robustness of the used model. The baseline model increases 5% to 10% F1 score on two adversarial SQuAD datasets without decreasing the F1 score on the original SQuAD dataset. Besides, by adding the NE filter, other existing models increases 5% F1 score on the adversarial datasets with less than 1% loss on the original one.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133144308","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, we propose a replay attack spoofing detection system for automatic speaker verification using multi-task learning of noise classes. We define the noise that is caused by the replay attack as replay noise. We explore the effectiveness of training a deep neural network simultaneously for replay attack spoofing detection and replay noise classification. The multi-task learning includes classifying the noise of playback devices, recording environments, and recording devices as well as the spoofing detection. Each of the three types of the noise classes also includes a genuine class. The experiment results on the version 1.0 of ASVspoof2017 datasets demonstrate that the performance of our proposed system is improved by 30% relatively on the evaluation set.
{"title":"Replay Spoofing Detection System for Automatic Speaker Verification Using Multi-Task Learning of Noise Classes","authors":"Hye-jin Shim, Jee-weon Jung, Hee-Soo Heo, Sung-Hyun Yoon, Ha-jin Yu","doi":"10.1109/TAAI.2018.00046","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00046","url":null,"abstract":"In this paper, we propose a replay attack spoofing detection system for automatic speaker verification using multi-task learning of noise classes. We define the noise that is caused by the replay attack as replay noise. We explore the effectiveness of training a deep neural network simultaneously for replay attack spoofing detection and replay noise classification. The multi-task learning includes classifying the noise of playback devices, recording environments, and recording devices as well as the spoofing detection. Each of the three types of the noise classes also includes a genuine class. The experiment results on the version 1.0 of ASVspoof2017 datasets demonstrate that the performance of our proposed system is improved by 30% relatively on the evaluation set.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-08-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134244051","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：