Choosing an appropriate distribution model for an e-commerce enterprise will undoubtedly greatly strengthen the core competitiveness of the enterprise. This paper aims to explore decision making of the best logistics distribution model to improve the distribution efficiency by taking the Dangdang E-commerce as an example. The research method used is Analytic Hierarchy Process (AHP), firstly to clarify the various impact indicators of the logistics distribution model selection, and then deeply analyze the factors affecting Dangdang's logistics distribution choices. The relative weights of each index are obtained through measurement, and then the scores are evaluated to obtain the priority of distribution model. The final delivery model provides Dangdang E-commerce recommendation.
{"title":"Using AHP to Choose the Best Logistics Distribution Model","authors":"M. Lai, Wu-Der Tsay, Zhen Wang","doi":"10.1109/TAAI.2018.00045","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00045","url":null,"abstract":"Choosing an appropriate distribution model for an e-commerce enterprise will undoubtedly greatly strengthen the core competitiveness of the enterprise. This paper aims to explore decision making of the best logistics distribution model to improve the distribution efficiency by taking the Dangdang E-commerce as an example. The research method used is Analytic Hierarchy Process (AHP), firstly to clarify the various impact indicators of the logistics distribution model selection, and then deeply analyze the factors affecting Dangdang's logistics distribution choices. The relative weights of each index are obtained through measurement, and then the scores are evaluated to obtain the priority of distribution model. The final delivery model provides Dangdang E-commerce recommendation.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115555504","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}
This study proposes a method to estimate players' skill using Computer Go. Computer Go has been considered one of the biggest challenges of artificial intelligence (AI) research. The AI of Go is based on the Monte Carlo tree search (MCTS) algorithm unlike the games of chess and shogi which are based on the game tree search using the evaluation function. Further, we apply the evaluation index used in the strength estimation method for shogi to the game of Go. We analyze the game records of KGS and YUUGEN-NO-MA with the evaluation index using the MCTS winning rate. It is concluded that stronger AI is necessary for identifying strength-estimating indicators.
本研究提出一种评估棋手围棋水平的方法。计算机围棋被认为是人工智能(AI)研究的最大挑战之一。围棋的人工智能是基于蒙特卡洛树搜索(Monte Carlo tree search, MCTS)算法，而国际象棋和将军棋则是基于使用评价函数的游戏树搜索。在此基础上，将将棋强度估计方法中的评价指标应用到围棋中。采用MCTS胜率评价指标对KGS和YUUGEN-NO-MA的比赛记录进行了分析。结论是需要更强的人工智能来识别强度估计指标。
{"title":"Examination of Indicators for Estimating Players' Strength by Using Computer Go","authors":"Yuuto Kosaka, Takeshi Ito","doi":"10.1109/TAAI.2018.00030","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00030","url":null,"abstract":"This study proposes a method to estimate players' skill using Computer Go. Computer Go has been considered one of the biggest challenges of artificial intelligence (AI) research. The AI of Go is based on the Monte Carlo tree search (MCTS) algorithm unlike the games of chess and shogi which are based on the game tree search using the evaluation function. Further, we apply the evaluation index used in the strength estimation method for shogi to the game of Go. We analyze the game records of KGS and YUUGEN-NO-MA with the evaluation index using the MCTS winning rate. It is concluded that stronger AI is necessary for identifying strength-estimating indicators.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127638284","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}
To reduce the number of accidents, this thesis proposes a vehicle path prediction system to predict the future direction when a vehicle is about to cross an intersection. The GPS sensor is used to collect the dataset of vehicle trajectories at intersections. The trend of vehicle movements are derived from the heading in the trajectories, which is then combined with the vehicle speed to generate training data. In our path prediction algorithm, two ensemble learning algorithms, i.e., Random Forests and AdaBoost, are adopted for model training. The experiment results indicate that the Random Forest algorithm exhibits the best performance, and the Adaboost algorithm performs better than the base learner (i.e., Decision Tree).
{"title":"Machine Learning Based Path Prediction System - Adapting One Model for All Intersections","authors":"Kai-Qi Huang, Min-Te Sun","doi":"10.1109/TAAI.2018.00023","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00023","url":null,"abstract":"To reduce the number of accidents, this thesis proposes a vehicle path prediction system to predict the future direction when a vehicle is about to cross an intersection. The GPS sensor is used to collect the dataset of vehicle trajectories at intersections. The trend of vehicle movements are derived from the heading in the trajectories, which is then combined with the vehicle speed to generate training data. In our path prediction algorithm, two ensemble learning algorithms, i.e., Random Forests and AdaBoost, are adopted for model training. The experiment results indicate that the Random Forest algorithm exhibits the best performance, and the Adaboost algorithm performs better than the base learner (i.e., Decision Tree).","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126828032","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}
Pedestrian attribute recognition has many applications in surveillance and attribute based query, tracking, and person re-identification. The recent trend in deep-learning based pedestrian attribute recognition is to use a shared CNN backbone for feature extraction and multiple subsequent branches for the individual branches. While this allows the end-to-end learning to simultaneously recognize multiple attributes, the data imbalance problem of most attributes becomes a challenge that has not been studied sufficiently for this application. This paper presents studies on how the cost adjustment method affects several common evaluation metrics. We also propose a two-stage training procedure, where an additional fine-tuning stage on the classifier layers only with class-balanced data is shown to improve recognition performances.
{"title":"On the Effect of Data Imbalance for Multi-Label Pedestrian Attribute Recognition","authors":"T. Wang, Kai-Chen Shu, Chia-Hao Chang, Yi-Fu Chen","doi":"10.1109/TAAI.2018.00025","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00025","url":null,"abstract":"Pedestrian attribute recognition has many applications in surveillance and attribute based query, tracking, and person re-identification. The recent trend in deep-learning based pedestrian attribute recognition is to use a shared CNN backbone for feature extraction and multiple subsequent branches for the individual branches. While this allows the end-to-end learning to simultaneously recognize multiple attributes, the data imbalance problem of most attributes becomes a challenge that has not been studied sufficiently for this application. This paper presents studies on how the cost adjustment method affects several common evaluation metrics. We also propose a two-stage training procedure, where an additional fine-tuning stage on the classifier layers only with class-balanced data is shown to improve recognition performances.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126192970","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}
Target-In this paper, we focus on using medicine for patients who have cardiac arrest then must have to do Cardiopulmonary Resuscitation (CPR). We want to know the medicine influence in predicting state of an illness deterioration. Therefore, we proposes a Medication for Cardiac Arrest Early Warning System (MCAEWS). It's not only assist physicians to early diagnose of an illness and immediately warning, but also increase sensitivity, decrease false positive rate and mortality rate. The most important role is greatly improve medical quality. Methods-In this study, the data is from the emergency department of National Taiwan University Hospital (NTUH). It is from January 2014 to December 2015. The patients who stayed in the emergency detention area for more than six hours during this two years. The patients were included in the retrospective cohort study. To comparative measures for the machine learning models, we used such as the Area Under the Receiver Operating Characteristic Curve (AUROC) and the Area under the Precision-Recall Curve (AUPRC). Results-The data were analyzed for CPR and non-CPR groups respectively. Furthermore, we evaluated sensitivity and specificity. The Random Forest Algorithm (AUC: 0.98; AUP: 0.23) compare with others such as Logistic Regression Algorithm (AUC: 0.94; AUP: 0.13), Decision Tree (AUC: 0.97; AUP: 0.05), and Extreme Random Tree (AUC: 0.91; AUP: 0.08), it was significantly high performance. Conclusion-Increasing the drug factors in vital signs, that it effectively improved the accuracy of predicting cardiac arrest. The results of this study, it's help for emergency clinical Physicians and hospital quality management will validly solve clinical medical resource allocation issues and improve medical quality through decision support systems.
{"title":"Using Machine Learning Algorithms in Medication for Cardiac Arrest Early Warning System Construction and Forecasting","authors":"Hsiao-ko Chang, Cheng-Tse Wu, Ji-Han Liu, J. Jang","doi":"10.1109/TAAI.2018.00010","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00010","url":null,"abstract":"Target-In this paper, we focus on using medicine for patients who have cardiac arrest then must have to do Cardiopulmonary Resuscitation (CPR). We want to know the medicine influence in predicting state of an illness deterioration. Therefore, we proposes a Medication for Cardiac Arrest Early Warning System (MCAEWS). It's not only assist physicians to early diagnose of an illness and immediately warning, but also increase sensitivity, decrease false positive rate and mortality rate. The most important role is greatly improve medical quality. Methods-In this study, the data is from the emergency department of National Taiwan University Hospital (NTUH). It is from January 2014 to December 2015. The patients who stayed in the emergency detention area for more than six hours during this two years. The patients were included in the retrospective cohort study. To comparative measures for the machine learning models, we used such as the Area Under the Receiver Operating Characteristic Curve (AUROC) and the Area under the Precision-Recall Curve (AUPRC). Results-The data were analyzed for CPR and non-CPR groups respectively. Furthermore, we evaluated sensitivity and specificity. The Random Forest Algorithm (AUC: 0.98; AUP: 0.23) compare with others such as Logistic Regression Algorithm (AUC: 0.94; AUP: 0.13), Decision Tree (AUC: 0.97; AUP: 0.05), and Extreme Random Tree (AUC: 0.91; AUP: 0.08), it was significantly high performance. Conclusion-Increasing the drug factors in vital signs, that it effectively improved the accuracy of predicting cardiac arrest. The results of this study, it's help for emergency clinical Physicians and hospital quality management will validly solve clinical medical resource allocation issues and improve medical quality through decision support systems.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115323145","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 problem of air pollution has become a serious issue in developed as well as developing countries. Unfortunately, most of the current solutions are not very effective and this makes it important to have an efficient early warning system for monitoring and forecasting air quality. Our main focus is to build a real-time forecasting system with high accuracy, and deploy it in Taiwan. In this paper, we propose a forecast method called Adaptive Iterative Forecast (AIF), which can predict the value of PM2.5 for the next few hours (by linear programming, normalization and time-series) based on the trend of historical data. The goal of this research is to develop an efficient and accurate forecast model. Through various comparative analyses, we have proved that our model can achieve significant results. Based on the results, we have also built a real-time forecasting system which allows the users to stay aware of the air quality and plan their day to day life.
{"title":"A Fast PM2.5 Forecast Approach Based on Time-Series Data Analysis, Regression and Regularization","authors":"Cyuan-Heng Luo, Hsuan Yang, Li-Pang Huang, Sachit Mahajan, Ling-Jyh Chen","doi":"10.1109/TAAI.2018.00026","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00026","url":null,"abstract":"The problem of air pollution has become a serious issue in developed as well as developing countries. Unfortunately, most of the current solutions are not very effective and this makes it important to have an efficient early warning system for monitoring and forecasting air quality. Our main focus is to build a real-time forecasting system with high accuracy, and deploy it in Taiwan. In this paper, we propose a forecast method called Adaptive Iterative Forecast (AIF), which can predict the value of PM2.5 for the next few hours (by linear programming, normalization and time-series) based on the trend of historical data. The goal of this research is to develop an efficient and accurate forecast model. Through various comparative analyses, we have proved that our model can achieve significant results. Based on the results, we have also built a real-time forecasting system which allows the users to stay aware of the air quality and plan their day to day life.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123391095","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}
For the game of Go, Chess, and Shogi (Japanese Chess), deep neural networks (DNNs) have contributed to building accurate evaluation functions, and many studies have attempted to create the so-called value network, which predicts the reward of a given state. A recent study of the value network for the game of Go has shown that a two-headed neural network with two different objectives can be trained effectively and performs better than a single-headed network. One of the two heads is called a value head and the other head, the policy head, predicts the next move at a given state. This multitask training makes the network more robust and improves the generalization performance. In this paper, we show that a simple discriminator network is an alternative target of multitask learning. Compared to the existing deep neural network, our proposed network can be designed more easily because of its simple output. Our experimental results showed that our discriminative target also makes the learning stable and the evaluation function trained by our method is comparable to the training of existing studies in terms of predicting the next move and playing strength.
{"title":"Alternative Multitask Training for Evaluation Functions in Game of Go","authors":"Yusaku Mandai, Tomoyuki Kaneko","doi":"10.1109/TAAI.2018.00037","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00037","url":null,"abstract":"For the game of Go, Chess, and Shogi (Japanese Chess), deep neural networks (DNNs) have contributed to building accurate evaluation functions, and many studies have attempted to create the so-called value network, which predicts the reward of a given state. A recent study of the value network for the game of Go has shown that a two-headed neural network with two different objectives can be trained effectively and performs better than a single-headed network. One of the two heads is called a value head and the other head, the policy head, predicts the next move at a given state. This multitask training makes the network more robust and improves the generalization performance. In this paper, we show that a simple discriminator network is an alternative target of multitask learning. Compared to the existing deep neural network, our proposed network can be designed more easily because of its simple output. Our experimental results showed that our discriminative target also makes the learning stable and the evaluation function trained by our method is comparable to the training of existing studies in terms of predicting the next move and playing strength.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122689271","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 AlphaZero algorithm, developed by DeepMind, achieved superhuman levels of play in the games of chess, shogi, and Go, by learning without domain-specific knowledge except game rules. This paper investigates whether the algorithm can also learn theoretical values and optimal plays for non-deterministic games. Since the theoretical values of such games are expected win rates, not a simple win, loss, or draw, it is worthy investigating the ability of the AlphaZero algorithm to approximate expected win rates of positions. This paper also studies how the algorithm is influenced by a set of hyper-parameters. The tested non-deterministic game is a reduced and solved version of Chinese dark chess (CDC), called 2×4 CDC. The experiments show that the AlphaZero algorithm converges nearly to the theoretical values and the optimal plays in many of the settings of the hyper-parameters. To our knowledge, this is the first research paper that applies the AlphaZero algorithm to non-deterministic games.
{"title":"AlphaZero for a Non-Deterministic Game","authors":"Chu-Hsuan Hsueh, I-Chen Wu, Jr-Chang Chen, T. Hsu","doi":"10.1109/TAAI.2018.00034","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00034","url":null,"abstract":"The AlphaZero algorithm, developed by DeepMind, achieved superhuman levels of play in the games of chess, shogi, and Go, by learning without domain-specific knowledge except game rules. This paper investigates whether the algorithm can also learn theoretical values and optimal plays for non-deterministic games. Since the theoretical values of such games are expected win rates, not a simple win, loss, or draw, it is worthy investigating the ability of the AlphaZero algorithm to approximate expected win rates of positions. This paper also studies how the algorithm is influenced by a set of hyper-parameters. The tested non-deterministic game is a reduced and solved version of Chinese dark chess (CDC), called 2×4 CDC. The experiments show that the AlphaZero algorithm converges nearly to the theoretical values and the optimal plays in many of the settings of the hyper-parameters. To our knowledge, this is the first research paper that applies the AlphaZero algorithm to non-deterministic games.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124619994","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}
Spontaneous Reporting Systems (SRSs) refer to systems used to collect voluntary reporting of adverse drug events (ADEs), which usually contain sensitive personal privacy information. Although many scholars have proposed various privacy protection models, they overlooked characteristics of SRS data. We previously have proposed a feasible privacy model and anonymization method dedicate to SRS data. However, this method is only applicable to complete data, not considering the fact that SRS data contain a lot of missing data. In this paper, we propose a new privacy model Closed MS(k, θ*)-bounding and a new anonymization method, Closed-MSpartition, to process SRS data with missing values. We used US FDA's FAERS data to evaluate our proposed method from the aspects of information loss, privacy risk, and data utility. The results show that our proposed new method can effectively prevent attackers from learning personal privacy without sacrificing data quality and utility.
{"title":"Privacy-Preserving SRS Data Anonymization by Incorporating Missing Values","authors":"Wen-Yang Lin, Kuang-Yung Hsu, Zih-Xun Shen","doi":"10.1109/TAAI.2018.00032","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00032","url":null,"abstract":"Spontaneous Reporting Systems (SRSs) refer to systems used to collect voluntary reporting of adverse drug events (ADEs), which usually contain sensitive personal privacy information. Although many scholars have proposed various privacy protection models, they overlooked characteristics of SRS data. We previously have proposed a feasible privacy model and anonymization method dedicate to SRS data. However, this method is only applicable to complete data, not considering the fact that SRS data contain a lot of missing data. In this paper, we propose a new privacy model Closed MS(k, θ*)-bounding and a new anonymization method, Closed-MSpartition, to process SRS data with missing values. We used US FDA's FAERS data to evaluate our proposed method from the aspects of information loss, privacy risk, and data utility. The results show that our proposed new method can effectively prevent attackers from learning personal privacy without sacrificing data quality and utility.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127382385","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}
As meta-heuristics to solve large-scale constraint satisfaction problems (CSPs), ant colony optimization (ACO) has recently been drawing attentions. In most of algorithms based on ACO, candidate assignments are constructed by taking account of data called pheromone graph. The pheromone graph is updated getting positive feedbacks from candidate assignments with the least number of constraint violations. However, it might be easy to get stuck in locally optimal solutions considering only a single perspective. In this paper, we propose a method that adopting new pheromone graph in addition to the original pheromone graph. This new pheromone graph is updated getting negative feedback from candidate assignments with the greatest number of constraint violations. This new pheromone graph, called a negative pheromone graph, is updated getting negative feedback from candidate assignments with the largest number of constraint violations. Also, the standard pheromone graph is updated by considering negative pheromones as well. By using pheromones updated from two perspectives, more effective search can be conducted. Moreover, in this paper, we conducted experiments on graph coloring problems. Graph coloring problem is one of CSPs. We demnonstrated that our model, which is applied to the cunning ant system, can be effective than other ACO-based methods for large-scale and hard graph coloring problems whose instance appears in the phase transition region.
{"title":"Ant Colony Optimization with Negative Feedback for Solving Constraint Satisfaction Problems","authors":"Takuya Masukane, Kazunori Mizuno, Hirotoshi Shinohara","doi":"10.1109/TAAI.2018.00041","DOIUrl":"https://doi.org/10.1109/TAAI.2018.00041","url":null,"abstract":"As meta-heuristics to solve large-scale constraint satisfaction problems (CSPs), ant colony optimization (ACO) has recently been drawing attentions. In most of algorithms based on ACO, candidate assignments are constructed by taking account of data called pheromone graph. The pheromone graph is updated getting positive feedbacks from candidate assignments with the least number of constraint violations. However, it might be easy to get stuck in locally optimal solutions considering only a single perspective. In this paper, we propose a method that adopting new pheromone graph in addition to the original pheromone graph. This new pheromone graph is updated getting negative feedback from candidate assignments with the greatest number of constraint violations. This new pheromone graph, called a negative pheromone graph, is updated getting negative feedback from candidate assignments with the largest number of constraint violations. Also, the standard pheromone graph is updated by considering negative pheromones as well. By using pheromones updated from two perspectives, more effective search can be conducted. Moreover, in this paper, we conducted experiments on graph coloring problems. Graph coloring problem is one of CSPs. We demnonstrated that our model, which is applied to the cunning ant system, can be effective than other ACO-based methods for large-scale and hard graph coloring problems whose instance appears in the phase transition region.","PeriodicalId":211734,"journal":{"name":"2018 Conference on Technologies and Applications of Artificial Intelligence (TAAI)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114176045","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：