Pub Date : 2020-01-01DOI: 10.4236/jilsa.2020.123004
Abdulai Ayirebi Ankrah, J. Kimotho, O. Muvengei
Transportation of freight and passengers by train is one of the oldest types of transport, and has now taken root in most of the developing countries especially in Africa. Recently, with the advent and development of high-speed trains, continuous monitoring of the railway vehicle suspension is of significant importance. For this reason, railway vehicles should be monitored continuously to avoid catastrophic events, ensure comfort, safety, and also improved performance while reducing life cycle costs. The suspension system is a very important part of the railway vehicle which supports the car-body and the bogie, isolates the forces generated by the track unevenness at the wheels and also controls the attitude of the car-body with respect to the track surface for ride comfort. Its reliability is directly related to the vehicle safety. The railway vehicle suspension often develops faults; worn springs and dampers in the primary and secondary suspension. To avoid a complete system failure, early detection of fault in the suspension of trains is of high importance. The main contribution of the research work is the prediction of faulty regimes of a railway vehicle suspension based on a hybrid model. The hybrid model framework is in four folds; first, modeling of vehicle suspension system to generate vertical acceleration of the railway vehicle, parameter estimation or identification was performed to obtain the nominal parameter values of the vehicle suspension system based on the measured data in the second fold, furthermore, a supervised machine learning model was built to predict faulty and healthy state of the suspension system components (damage scenarios) based on support vector machine (SVM) and lastly, the development of a new SVM model with the damage scenarios to predict faults on the test data. The level of degradation at which the spring and damper becomes faulty for both primary and secondary suspension system was determined. The spring and damper becomes faulty when the nominal values degrade by 50% and 40% and 30% and 40% for the secondary and primary suspension system respectively. The proposed model was able to predict faulty components with an accuracy of 0.844 for the primary and secondary suspension system.
{"title":"Fusion of Model-Based and Data Driven Based Fault Diagnostic Methods for Railway Vehicle Suspension","authors":"Abdulai Ayirebi Ankrah, J. Kimotho, O. Muvengei","doi":"10.4236/jilsa.2020.123004","DOIUrl":"https://doi.org/10.4236/jilsa.2020.123004","url":null,"abstract":"Transportation of freight and passengers by train is one of the oldest types of transport, and has now taken root in most of the developing countries especially in Africa. Recently, with the advent and development of high-speed trains, continuous monitoring of the railway vehicle suspension is of significant importance. For this reason, railway vehicles should be monitored continuously to avoid catastrophic events, ensure comfort, safety, and also improved performance while reducing life cycle costs. The suspension system is a very important part of the railway vehicle which supports the car-body and the bogie, isolates the forces generated by the track unevenness at the wheels and also controls the attitude of the car-body with respect to the track surface for ride comfort. Its reliability is directly related to the vehicle safety. The railway vehicle suspension often develops faults; worn springs and dampers in the primary and secondary suspension. To avoid a complete system failure, early detection of fault in the suspension of trains is of high importance. The main contribution of the research work is the prediction of faulty regimes of a railway vehicle suspension based on a hybrid model. The hybrid model framework is in four folds; first, modeling of vehicle suspension system to generate vertical acceleration of the railway vehicle, parameter estimation or identification was performed to obtain the nominal parameter values of the vehicle suspension system based on the measured data in the second fold, furthermore, a supervised machine learning model was built to predict faulty and healthy state of the suspension system components (damage scenarios) based on support vector machine (SVM) and lastly, the development of a new SVM model with the damage scenarios to predict faults on the test data. The level of degradation at which the spring and damper becomes faulty for both primary and secondary suspension system was determined. The spring and damper becomes faulty when the nominal values degrade by 50% and 40% and 30% and 40% for the secondary and primary suspension system respectively. The proposed model was able to predict faulty components with an accuracy of 0.844 for the primary and secondary suspension system.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"70330525","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}
Pub Date : 2019-09-30DOI: 10.4236/jilsa.2019.114004
V. Shats
This paper proposes two new algorithms for classifying objects with categorical attributes. These algorithms are derived from the assumption that the attributes of different object classes have different probability distributions. One algorithm classifies objects based on the distribution of the attribute frequencies, and the other classifies objects based on the distribution of the pairwise attribute frequencies described using a matrix of pairwise frequencies. Both algorithms are based on the method of invariants, which offers the simplest dependencies for estimating the probabilities of objects in each class by an average frequency of their attributes. The estimated object class corresponds to the maximum probability. This method reflects the sensory process models of animals and is aimed at recognizing an object class by searching for a prototype in information accumulated in the brain. Because these matrices may be sparse, the solution cannot be determined for some objects. For these objects, an analog of the k-nearest neighbors method is provided in which for each attribute value, the class to which the majority of the k-nearest objects in the training sample belong is determined, and the most likely class value is calculated. The efficiencies of these two algorithms were confirmed on five databases.
{"title":"On the Matrices of Pairwise Frequencies of Categorical Attributes for Objects Classification","authors":"V. Shats","doi":"10.4236/jilsa.2019.114004","DOIUrl":"https://doi.org/10.4236/jilsa.2019.114004","url":null,"abstract":"This paper proposes two new algorithms for classifying objects with categorical attributes. These algorithms are derived from the assumption that the attributes of different object classes have different probability distributions. One algorithm classifies objects based on the distribution of the attribute frequencies, and the other classifies objects based on the distribution of the pairwise attribute frequencies described using a matrix of pairwise frequencies. Both algorithms are based on the method of invariants, which offers the simplest dependencies for estimating the probabilities of objects in each class by an average frequency of their attributes. The estimated object class corresponds to the maximum probability. This method reflects the sensory process models of animals and is aimed at recognizing an object class by searching for a prototype in information accumulated in the brain. Because these matrices may be sparse, the solution cannot be determined for some objects. For these objects, an analog of the k-nearest neighbors method is provided in which for each attribute value, the class to which the majority of the k-nearest objects in the training sample belong is determined, and the most likely class value is calculated. The efficiencies of these two algorithms were confirmed on five databases.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46541922","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}
Pub Date : 2019-08-14DOI: 10.4236/JILSA.2019.113003
Jiaxin Gao, Zirui Zhou, Jiangshan Ai, Bingxin Xia, Stephen Coggeshall
Credit card fraud is a wide-ranging issue for financial institutions, involving theft and fraud committed using a payment card. In this paper, we explore the application of linear and nonlinear statistical modeling and machine learning models on real credit card transaction data. The models built are supervised fraud models that attempt to identify which transactions are most likely fraudulent. We discuss the processes of data exploration, data cleaning, variable creation, feature selection, model algorithms, and results. Five different supervised models are explored and compared including logistic regression, neural networks, random forest, boosted tree and support vector machines. The boosted tree model shows the best fraud detection result (FDR = 49.83%) for this particular data set. The resulting model can be utilized in a credit card fraud detection system. A similar model development process can be performed in related business domains such as insurance and telecommunications, to avoid or detect fraudulent activity.
{"title":"Predicting Credit Card Transaction Fraud Using Machine Learning Algorithms","authors":"Jiaxin Gao, Zirui Zhou, Jiangshan Ai, Bingxin Xia, Stephen Coggeshall","doi":"10.4236/JILSA.2019.113003","DOIUrl":"https://doi.org/10.4236/JILSA.2019.113003","url":null,"abstract":"Credit card fraud is a wide-ranging issue for financial institutions, involving theft and fraud committed using a payment card. In this paper, we explore the application of linear and nonlinear statistical modeling and machine learning models on real credit card transaction data. The models built are supervised fraud models that attempt to identify which transactions are most likely fraudulent. We discuss the processes of data exploration, data cleaning, variable creation, feature selection, model algorithms, and results. Five different supervised models are explored and compared including logistic regression, neural networks, random forest, boosted tree and support vector machines. The boosted tree model shows the best fraud detection result (FDR = 49.83%) for this particular data set. The resulting model can be utilized in a credit card fraud detection system. A similar model development process can be performed in related business domains such as insurance and telecommunications, to avoid or detect fraudulent activity.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"70330911","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}
Pub Date : 2019-03-07DOI: 10.4236/JILSA.2019.112002
Xiyu Kang, Yiqi Wang, Yanrui Hu
Minimax algorithm and machine learning technologies have been studied for decades to reach an ideal optimization in game areas such as chess and backgammon. In these fields, several generations try to optimize the code for pruning and effectiveness of evaluation function. Thus, there are well-armed algorithms to deal with various sophisticated situations in gaming occasion. However, as a traditional zero-sum game, Connect-4 receives less attention compared with the other members of its zero-sum family using traditional minimax algorithm. In recent years, new generation of heuristics is created to address this problem based on research conclusions, expertise and gaming experiences. However, this paper mainly introduced a self-developed heuristics supported by well-demonstrated result from researches and our own experiences which fighting against the available version of Connect-4 system online. While most previous works focused on winning algorithms and knowledge based approaches, we complement these works with analysis of heuristics. We have conducted three experiments on the relationship among functionality, depth of searching and number of features and doing contrastive test with sample online. Different from the sample based on summarized experience and generalized features, our heuristics have a basic concentration on detailed connection between pieces on board. By analysing the winning percentages when our version fights against the online sample with different searching depths, we find that our heuristics with minimax algorithm is perfect on the early stages of the zero-sum game playing. Because some nodes in the game tree have no influence on the final decision of minimax algorithm, we use alpha-beta pruning to decrease the number of meaningless node which greatly increases the minimax efficiency. During the contrastive experiment with the online sample, this paper also verifies basic characters of the minimax algorithm including depths and quantity of features. According to the experiment, these two characters can both effect the decision for each step and none of them can be absolutely in charge. Besides, we also explore some potential future issues in Connect-4 game optimization such as precise adjustment on heuristic values and inefficiency pruning on the search tree.
{"title":"Research on Different Heuristics for Minimax Algorithm Insight from Connect-4 Game","authors":"Xiyu Kang, Yiqi Wang, Yanrui Hu","doi":"10.4236/JILSA.2019.112002","DOIUrl":"https://doi.org/10.4236/JILSA.2019.112002","url":null,"abstract":"Minimax algorithm and machine learning technologies have been studied for decades to reach an ideal optimization in game areas such as chess and backgammon. In these fields, several generations try to optimize the code for pruning and effectiveness of evaluation function. Thus, there are well-armed algorithms to deal with various sophisticated situations in gaming occasion. However, as a traditional zero-sum game, Connect-4 receives less attention compared with the other members of its zero-sum family using traditional minimax algorithm. In recent years, new generation of heuristics is created to address this problem based on research conclusions, expertise and gaming experiences. However, this paper mainly introduced a self-developed heuristics supported by well-demonstrated result from researches and our own experiences which fighting against the available version of Connect-4 system online. While most previous works focused on winning algorithms and knowledge based approaches, we complement these works with analysis of heuristics. We have conducted three experiments on the relationship among functionality, depth of searching and number of features and doing contrastive test with sample online. Different from the sample based on summarized experience and generalized features, our heuristics have a basic concentration on detailed connection between pieces on board. By analysing the winning percentages when our version fights against the online sample with different searching depths, we find that our heuristics with minimax algorithm is perfect on the early stages of the zero-sum game playing. Because some nodes in the game tree have no influence on the final decision of minimax algorithm, we use alpha-beta pruning to decrease the number of meaningless node which greatly increases the minimax efficiency. During the contrastive experiment with the online sample, this paper also verifies basic characters of the minimax algorithm including depths and quantity of features. According to the experiment, these two characters can both effect the decision for each step and none of them can be absolutely in charge. Besides, we also explore some potential future issues in Connect-4 game optimization such as precise adjustment on heuristic values and inefficiency pruning on the search tree.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46439318","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}
Pub Date : 2019-01-18DOI: 10.4236/JILSA.2019.111001
A. Khedher, I. Jraidi, C. Frasson
Monitoring students’ level of engagement during learning activities is an important challenge in the development of tutoring interventions. In this paper, we explore the feasibility of using electroencephalographic signals (EEG) as a tool to monitor the mental engagement index of novice medicine students during a reasoning process. More precisely, the objectives were first, to track students’ mental engagement evolution in order to investigate whether there were particular sections within the learning environment that aroused the highest engagement level among the students, and, if so, did these sections have an impact on learners’ performance. Experimental analyses showed the same trends in the different resolution phases as well as across the different regions of the environments. However, we noticed a higher engagement index during the treatment identification phase since it aroused more mental effort. Moreover statistically significant effects were found between mental engagement and students’ performance.
{"title":"Tracking Students’ Mental Engagement Using EEG Signals during an Interaction with a Virtual Learning Environment","authors":"A. Khedher, I. Jraidi, C. Frasson","doi":"10.4236/JILSA.2019.111001","DOIUrl":"https://doi.org/10.4236/JILSA.2019.111001","url":null,"abstract":"Monitoring students’ level of engagement during learning activities is an important challenge in the development of tutoring interventions. In this paper, we explore the feasibility of using electroencephalographic signals (EEG) as a tool to monitor the mental engagement index of novice medicine students during a reasoning process. More precisely, the objectives were first, to track students’ mental engagement evolution in order to investigate whether there were particular sections within the learning environment that aroused the highest engagement level among the students, and, if so, did these sections have an impact on learners’ performance. Experimental analyses showed the same trends in the different resolution phases as well as across the different regions of the environments. However, we noticed a higher engagement index during the treatment identification phase since it aroused more mental effort. Moreover statistically significant effects were found between mental engagement and students’ performance.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2019-01-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"41749449","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}
Pub Date : 2018-10-24DOI: 10.4236/JILSA.2018.104008
Ziming Chi, Bing Zhang
In this paper we employ an improved Siamese neural network to assess the semantic similarity between sentences. Our model implements the function of inputting two sentences to obtain the similarity score. We design our model based on the Siamese network using deep Long Short-Term Memory (LSTM) Network. And we add the special attention mechanism to let the model give different words different attention while modeling sentences. The fully-connected layer is proposed to measure the complex sentence representations. Our results show that the accuracy is better than the baseline in 2016. Furthermore, it is showed that the model has the ability to model the sequence order, distribute reasonable attention and extract meanings of a sentence in different dimensions.
{"title":"A Sentence Similarity Estimation Method Based on Improved Siamese Network","authors":"Ziming Chi, Bing Zhang","doi":"10.4236/JILSA.2018.104008","DOIUrl":"https://doi.org/10.4236/JILSA.2018.104008","url":null,"abstract":"In this paper we employ an improved Siamese neural network to assess the semantic similarity between sentences. Our model implements the function of inputting two sentences to obtain the similarity score. We design our model based on the Siamese network using deep Long Short-Term Memory (LSTM) Network. And we add the special attention mechanism to let the model give different words different attention while modeling sentences. The fully-connected layer is proposed to measure the complex sentence representations. Our results show that the accuracy is better than the baseline in 2016. Furthermore, it is showed that the model has the ability to model the sequence order, distribute reasonable attention and extract meanings of a sentence in different dimensions.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-10-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"41623827","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}
Pub Date : 2018-08-15DOI: 10.4236/JILSA.2018.103007
Tianquan Feng, Qingrong Chen, M. Yi
We theoretically investigate the collective response of an ensemble of leaky integrate-and-fire neuron units to a noisy periodic signal by including local spatially correlated noise. By using the linear response theory, we obtained the analytic expression of signal-to-noise ratio (SNR). Numerical simulation results show that the rms amplitude of internal noise can be increased up to an optimal value where the output SNR reaches a maximum value. Due to the existence of the local spatially correlated noise in the units of the ensemble, the SNR gain of the collective ensemble response can exceed unity and can be optimized when the nearest-neighborhood correlation is negative. This nonlinear collective phenomenon of SNR gain amplification in an ensemble of leaky integrate-and-fire neuron units can be related to the array stochastic resonance (SR) phenomenon. Furthermore, we also show that the SNR gain can also be optimized by tuning the number of neuron units, frequency and amplitude of the weak periodic signal. The present study illustrates the potential to utilize the local spatially correlation noise and the number of ensemble units for optimizing the collective response of the neuron to inputs, as well as a guidance in the design of information processing devices to weak signal detection.
{"title":"Local Correlated Noise Improvement of Signal-to-Noise Ratio Gain in an Ensemble of Noisy Neuron","authors":"Tianquan Feng, Qingrong Chen, M. Yi","doi":"10.4236/JILSA.2018.103007","DOIUrl":"https://doi.org/10.4236/JILSA.2018.103007","url":null,"abstract":"We theoretically investigate the collective response of an ensemble of leaky integrate-and-fire neuron units to a noisy periodic signal by including local spatially correlated noise. By using the linear response theory, we obtained the analytic expression of signal-to-noise ratio (SNR). Numerical simulation results show that the rms amplitude of internal noise can be increased up to an optimal value where the output SNR reaches a maximum value. Due to the existence of the local spatially correlated noise in the units of the ensemble, the SNR gain of the collective ensemble response can exceed unity and can be optimized when the nearest-neighborhood correlation is negative. This nonlinear collective phenomenon of SNR gain amplification in an ensemble of leaky integrate-and-fire neuron units can be related to the array stochastic resonance (SR) phenomenon. Furthermore, we also show that the SNR gain can also be optimized by tuning the number of neuron units, frequency and amplitude of the weak periodic signal. The present study illustrates the potential to utilize the local spatially correlation noise and the number of ensemble units for optimizing the collective response of the neuron to inputs, as well as a guidance in the design of information processing devices to weak signal detection.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47022602","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}
Pub Date : 2018-08-03DOI: 10.4236/JILSA.2018.103006
R. Ghali, H. Abdessalem, C. Frasson, R. Nkambou
Gifted students have different ways of learning. They are characterized by a fitful level of attention and intuitive reasoning. In order to distinguish gifted students from normal students, we conducted an experiment with 17 pupils, willing participants in this study. We collected different types of data (gender, age, performance, initial average in math and EEG mental states) in a web platform called NetMath intending for the learning of mathematics. We selected ten tasks divided into three difficulty levels (easy, medium and hard). Participants were invited to respond to top-level exercises on the four basic operations in decimals. Our first results confirmed that the student’s performance has no relation with age. A younger 9-year-old student achieved a higher score than the group with an average of 68.18%. This student can be considered as a gifted one. The gifted students can be also characterized by a mean value of attention (around 60%). They also can be defined by slightly weaker values of their mental states of attention and workload in comparison with the weak pupils.
{"title":"Identifying Brain Characteristics of Bright Students","authors":"R. Ghali, H. Abdessalem, C. Frasson, R. Nkambou","doi":"10.4236/JILSA.2018.103006","DOIUrl":"https://doi.org/10.4236/JILSA.2018.103006","url":null,"abstract":"Gifted students have different ways of learning. They are characterized by a fitful level of attention and intuitive reasoning. In order to distinguish gifted students from normal students, we conducted an experiment with 17 pupils, willing participants in this study. We collected different types of data (gender, age, performance, initial average in math and EEG mental states) in a web platform called NetMath intending for the learning of mathematics. We selected ten tasks divided into three difficulty levels (easy, medium and hard). Participants were invited to respond to top-level exercises on the four basic operations in decimals. Our first results confirmed that the student’s performance has no relation with age. A younger 9-year-old student achieved a higher score than the group with an average of 68.18%. This student can be considered as a gifted one. The gifted students can be also characterized by a mean value of attention (around 60%). They also can be defined by slightly weaker values of their mental states of attention and workload in comparison with the weak pupils.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44122216","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}
Pub Date : 2018-08-03DOI: 10.4236/jilsa.2018.103005
V. Shats
The present paper solves the training problem that comprises the initial phases of the classification problem using the data matrix invariant method. The method is reduced to an approximate “slicing” of the information contained in the problem, which leads to its structuring. According to this method, the values of each feature are divided into an equal number of intervals, and lists of objects falling into these intervals are constructed. Objects are identified by a set of numbers of intervals, i.e., indices, for each feature. Assuming that the feature values within any interval are approximately the same, we calculate frequency features for objects of different classes that are equal to the frequencies of the corresponding indices. These features allow us to determine the frequency of any object class as the sum of the frequencies of the indices. For any number of intervals, the maximum frequency corresponds to a class object. If the features do not contain repeated values, the error rate of training tends to zero for an infinite number of intervals. If this condition is not fulfilled, a preliminary randomization of the features should be carried out.
{"title":"Error-Free Training via Information Structuring in the Classification Problem","authors":"V. Shats","doi":"10.4236/jilsa.2018.103005","DOIUrl":"https://doi.org/10.4236/jilsa.2018.103005","url":null,"abstract":"The present paper solves the training problem that comprises the initial phases of the classification problem using the data matrix invariant method. The method is reduced to an approximate “slicing” of the information contained in the problem, which leads to its structuring. According to this method, the values of each feature are divided into an equal number of intervals, and lists of objects falling into these intervals are constructed. Objects are identified by a set of numbers of intervals, i.e., indices, for each feature. Assuming that the feature values within any interval are approximately the same, we calculate frequency features for objects of different classes that are equal to the frequencies of the corresponding indices. These features allow us to determine the frequency of any object class as the sum of the frequencies of the indices. For any number of intervals, the maximum frequency corresponds to a class object. If the features do not contain repeated values, the error rate of training tends to zero for an infinite number of intervals. If this condition is not fulfilled, a preliminary randomization of the features should be carried out.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-08-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"45271409","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}
Pub Date : 2018-05-07DOI: 10.4236/jilsa.2018.102004
William Li
Teaching computer programs to play games through machine learning has been an important way to achieve better artificial intelligence (AI) in a variety of real-world applications. Monte Carlo Tree Search (MCTS) is one of the key AI techniques developed recently that enabled AlphaGo to defeat a legendary professional Go player. What makes MCTS particularly attractive is that it only understands the basic rules of the game and does not rely on expert-level knowledge. Researchers thus expect that MCTS can be applied to other complex AI problems where domain-specific expert-level knowledge is not yet available. So far there are very few analytic studies in the literature. In this paper, our goal is to develop analytic studies of MCTS to build a more fundamental understanding of the algorithms and their applicability in complex AI problems. We start with a simple version of MCTS, called random playout search (RPS), to play Tic-Tac-Toe, and find that RPS may fail to discover the correct moves even in a very simple game position of Tic-Tac-Toe. Both the probability analysis and simulation have confirmed our discovery. We continue our studies with the full version of MCTS to play Gomoku and find that while MCTS has shown great success in playing more sophisticated games like Go, it is not effective to address the problem of sudden death/win. The main reason that MCTS often fails to detect sudden death/win lies in the random playout search nature of MCTS, which leads to prediction distortion. Therefore, although MCTS in theory converges to the optimal minimax search, with real world computational resource constraints, MCTS has to rely on RPS as an important step in its search process, therefore suffering from the same fundamental prediction distortion problem as RPS does. By examining the detailed statistics of the scores in MCTS, we investigate a variety of scenarios where MCTS fails to detect sudden death/win. Finally, we propose an improved MCTS algorithm by incorporating minimax search to overcome prediction distortion. Our simulation has confirmed the effectiveness of the proposed algorithm. We provide an estimate of the additional computational costs of this new algorithm to detect sudden death/win and discuss heuristic strategies to further reduce the search complexity.
{"title":"Prediction Distortion in Monte Carlo Tree Search and an Improved Algorithm","authors":"William Li","doi":"10.4236/jilsa.2018.102004","DOIUrl":"https://doi.org/10.4236/jilsa.2018.102004","url":null,"abstract":"Teaching computer programs to play games through machine learning has been an important way to achieve better artificial intelligence (AI) in a variety of real-world applications. Monte Carlo Tree Search (MCTS) is one of the key AI techniques developed recently that enabled AlphaGo to defeat a legendary professional Go player. What makes MCTS particularly attractive is that it only understands the basic rules of the game and does not rely on expert-level knowledge. Researchers thus expect that MCTS can be applied to other complex AI problems where domain-specific expert-level knowledge is not yet available. So far there are very few analytic studies in the literature. In this paper, our goal is to develop analytic studies of MCTS to build a more fundamental understanding of the algorithms and their applicability in complex AI problems. We start with a simple version of MCTS, called random playout search (RPS), to play Tic-Tac-Toe, and find that RPS may fail to discover the correct moves even in a very simple game position of Tic-Tac-Toe. Both the probability analysis and simulation have confirmed our discovery. We continue our studies with the full version of MCTS to play Gomoku and find that while MCTS has shown great success in playing more sophisticated games like Go, it is not effective to address the problem of sudden death/win. The main reason that MCTS often fails to detect sudden death/win lies in the random playout search nature of MCTS, which leads to prediction distortion. Therefore, although MCTS in theory converges to the optimal minimax search, with real world computational resource constraints, MCTS has to rely on RPS as an important step in its search process, therefore suffering from the same fundamental prediction distortion problem as RPS does. By examining the detailed statistics of the scores in MCTS, we investigate a variety of scenarios where MCTS fails to detect sudden death/win. Finally, we propose an improved MCTS algorithm by incorporating minimax search to overcome prediction distortion. Our simulation has confirmed the effectiveness of the proposed algorithm. We provide an estimate of the additional computational costs of this new algorithm to detect sudden death/win and discuss heuristic strategies to further reduce the search complexity.","PeriodicalId":69452,"journal":{"name":"智能学习系统与应用(英文)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2018-05-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48023206","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：