Pub Date : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492826
Kee-Hoon Kim, Sung-Bae Cho
Recently, ubiquitous computing and related sensor technology have significantly progressed. On the other side, the relationship between human emotion and sensory stimuli has been investigated. With this background, we propose sensory stimuli control system to adjust group emotion to given target emotion. Valence-arousal model was adapted for defining group emotion, and survey of 73-papers and onsite-investigation had done for domain knowledge. The proposed system is based on the partially observable Markov decision process to deal with the uncertain states of group emotion, and reinforcement learning approach to learn the criterion of decision in real time. To evaluate the proposed system, we collected 160-minutes data from kindergarten where the music and math classes are ongoing with 10 prescholers and 1 caregiver are participating. Our system produced 55.17% of accuracy, which outperfomed the original system by 15.51%p.
{"title":"A group emotion control system based on reinforcement learning","authors":"Kee-Hoon Kim, Sung-Bae Cho","doi":"10.1109/SOCPAR.2015.7492826","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492826","url":null,"abstract":"Recently, ubiquitous computing and related sensor technology have significantly progressed. On the other side, the relationship between human emotion and sensory stimuli has been investigated. With this background, we propose sensory stimuli control system to adjust group emotion to given target emotion. Valence-arousal model was adapted for defining group emotion, and survey of 73-papers and onsite-investigation had done for domain knowledge. The proposed system is based on the partially observable Markov decision process to deal with the uncertain states of group emotion, and reinforcement learning approach to learn the criterion of decision in real time. To evaluate the proposed system, we collected 160-minutes data from kindergarten where the music and math classes are ongoing with 10 prescholers and 1 caregiver are participating. Our system produced 55.17% of accuracy, which outperfomed the original system by 15.51%p.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"17 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120917793","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}
Differential Evolution (DE) is an evolutionary algorithm. DE has been successfully applied to optimization problems including non-linear, non-differentiable, non-convex and multi-modal functions. The performance of DE is affected by algorithm parameters such as a scaling factor F and a crossover rate CR. Many studies have been done to control the parameters adaptively. One of the most successful studies on parameter control is JADE. In JADE, two parameter values are generated according to a probability density function which is learned by the parameter values in success cases, where the child is better than the parent. In this study, landscape of an objective function is paid attention to in order to improve the performance of JADE. The efficiency and robustness of search process can be improved by detecting valleys and hills in search points and by adopting a small F for valley points and a large F for hill points because an optimal solution exists near valleys and far from hills in minimization problems. Valley points and hill points are detected by creating a proximity graph from search points and by selecting valley/hill points that are smaller/greater than neighbor points. The effect of the proposed method is shown by solving thirteen benchmark problems.
{"title":"Improving an adaptive differential evolution using hill-valley detection","authors":"T. Takahama, S. Sakai","doi":"10.3233/HIS-160220","DOIUrl":"https://doi.org/10.3233/HIS-160220","url":null,"abstract":"Differential Evolution (DE) is an evolutionary algorithm. DE has been successfully applied to optimization problems including non-linear, non-differentiable, non-convex and multi-modal functions. The performance of DE is affected by algorithm parameters such as a scaling factor F and a crossover rate CR. Many studies have been done to control the parameters adaptively. One of the most successful studies on parameter control is JADE. In JADE, two parameter values are generated according to a probability density function which is learned by the parameter values in success cases, where the child is better than the parent. In this study, landscape of an objective function is paid attention to in order to improve the performance of JADE. The efficiency and robustness of search process can be improved by detecting valleys and hills in search points and by adopting a small F for valley points and a large F for hill points because an optimal solution exists near valleys and far from hills in minimization problems. Valley points and hill points are detected by creating a proximity graph from search points and by selecting valley/hill points that are smaller/greater than neighbor points. The effect of the proposed method is shown by solving thirteen benchmark problems.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125563580","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492789
W. Homenda, A. Jastrzębska
In the paper we look closely at the issue of contaminated data sets, where apart from proper elements we may have garbage. In a typical scenario, further classification of such data sets is always negatively influenced by garbage elements. Ideally, we would like to remove them from the data set entirely. Garbage elements are called here foreign elements and the task of removing them from the data set is called rejection of foreign elements. The paper is devoted to comparison and analysis of three different models capable to perform classification with rejection of foreign elements. It shall be emphasized that all studied methods are based only on proper patterns and no knowledge about foreign elements is needed to construct them. Hence, the methods we study are truly general and could be applied in many ways and in many problems. The following classification/rejection architectures are considered: global, local, and embedded. We analyze their performance in two aspects: influence of rejection mechanisms on classification and the quality of rejection. Issues are addressed theoretically and empirically in a study of handwritten digits recognition. Results show that the local architecture and the embedded architecture are advantageous, in comparison to the global architecture.
{"title":"Global, local and embedded architectures for multiclass classification with foreign elements rejection: An overview","authors":"W. Homenda, A. Jastrzębska","doi":"10.1109/SOCPAR.2015.7492789","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492789","url":null,"abstract":"In the paper we look closely at the issue of contaminated data sets, where apart from proper elements we may have garbage. In a typical scenario, further classification of such data sets is always negatively influenced by garbage elements. Ideally, we would like to remove them from the data set entirely. Garbage elements are called here foreign elements and the task of removing them from the data set is called rejection of foreign elements. The paper is devoted to comparison and analysis of three different models capable to perform classification with rejection of foreign elements. It shall be emphasized that all studied methods are based only on proper patterns and no knowledge about foreign elements is needed to construct them. Hence, the methods we study are truly general and could be applied in many ways and in many problems. The following classification/rejection architectures are considered: global, local, and embedded. We analyze their performance in two aspects: influence of rejection mechanisms on classification and the quality of rejection. Issues are addressed theoretically and empirically in a study of handwritten digits recognition. Results show that the local architecture and the embedded architecture are advantageous, in comparison to the global architecture.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122633695","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492805
Ramazan Algin, A. F. Alkaya
In this study, we tackle the obstacle neutralization problem wherein an agent is supposed to find the shortest path from given points s to t in a mapped hazard field where there are N potential mine discs in the field. In this problem agent has neutralization capability but he/she can neutralize only limited number of discs (K). The neutralization number is limited because of a specific reason such as the load capacity of agent or vehicle. When a disk is neutralized its cost is added to the traversal length of path. This problem is a kind of shortest problem with source constraints and it is NP-Hard. In this study, three important swarm intelligence techniques, namely ant system, ant colony system and migrating birds optimization algorithms, are applied to solve the obstacle neutralization problem and computational research is conducted in order to reveal their performance. Our experiments suggest that the migrating birds optimization algorithm outperforms ant system and ant colony system whereas ant colony system is better than ant system.
{"title":"Solving the obstacle neutralization problem using swarm intelligence algorithms","authors":"Ramazan Algin, A. F. Alkaya","doi":"10.1109/SOCPAR.2015.7492805","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492805","url":null,"abstract":"In this study, we tackle the obstacle neutralization problem wherein an agent is supposed to find the shortest path from given points s to t in a mapped hazard field where there are N potential mine discs in the field. In this problem agent has neutralization capability but he/she can neutralize only limited number of discs (K). The neutralization number is limited because of a specific reason such as the load capacity of agent or vehicle. When a disk is neutralized its cost is added to the traversal length of path. This problem is a kind of shortest problem with source constraints and it is NP-Hard. In this study, three important swarm intelligence techniques, namely ant system, ant colony system and migrating birds optimization algorithms, are applied to solve the obstacle neutralization problem and computational research is conducted in order to reveal their performance. Our experiments suggest that the migrating birds optimization algorithm outperforms ant system and ant colony system whereas ant colony system is better than ant system.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114733968","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492801
Prarinya Siritanawan, K. Kotani
Typical facial expression recognition system in computer vision field usually learns and translates facial behaviors into emotional states directly based on the training data. Since our face are not limited by a small number of class labels. In order to explain more complex facial expressions, we proposed a novel action unit (AU) detector following the Ekman's Facial Action Coding System (FACS). Our AU detection system utilized the robust temporal features and a new architecture of classification methods based on discriminative Independent Component Analysis (ICA) with whitening process by Eigenspace Method based on Class features (EMC). Therefore we can objectively describe the subtle and complex facial expressions in the same standard in psychology studies. The experimental results show the higher performance of our proposed system comparing to our previous classification methods in the standard dataset.
{"title":"Facial action units detection by robust temporal features","authors":"Prarinya Siritanawan, K. Kotani","doi":"10.1109/SOCPAR.2015.7492801","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492801","url":null,"abstract":"Typical facial expression recognition system in computer vision field usually learns and translates facial behaviors into emotional states directly based on the training data. Since our face are not limited by a small number of class labels. In order to explain more complex facial expressions, we proposed a novel action unit (AU) detector following the Ekman's Facial Action Coding System (FACS). Our AU detection system utilized the robust temporal features and a new architecture of classification methods based on discriminative Independent Component Analysis (ICA) with whitening process by Eigenspace Method based on Class features (EMC). Therefore we can objectively describe the subtle and complex facial expressions in the same standard in psychology studies. The experimental results show the higher performance of our proposed system comparing to our previous classification methods in the standard dataset.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126429164","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492769
A. Ghazvini, M. Nazri, S. Abdullah, Md Nawawi Junoh, Zainal Abidin bin Kasim
In sphere of criminology, suspect prediction analysis has been the point of convergence for many researchers. The focus of this study is on three prime attributes of next serial suspect's biography including nationality, age and time. Generally, to prevent the uncertainty in dynamic systems by nonlinear methods, a predictor is required in Time Delay Neural Network (TDNN). However, existing TDNN with single activation function is less effective to predict labeled class due to lower accuracy. Poor approximation of smooth mapping in single hidden layer makes it less effective. This study aims to propose a combined transfer functions to improve Nonlinear Autoregressive Time Series for performance prediction with exogenous (external) input (NARX)'s by utilizing Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) algorithms. Consequently Hyperbolic Tangent Sigmoid (Tansig) and Radial Basis Function (RBF) are used in LM and SCG algorithms as bi-transfer functions for prediction of next suspect's biography in commercial serial case. The results of NARX model with combination of Tansig and RBF as two objective of transfer functions of LM and SCG, presented better performance for prediction of next serial crime suspect's biography in comparison to single activation function of Tansig and RBF.
{"title":"Biography commercial serial crime analysis using enhanced dynamic neural network","authors":"A. Ghazvini, M. Nazri, S. Abdullah, Md Nawawi Junoh, Zainal Abidin bin Kasim","doi":"10.1109/SOCPAR.2015.7492769","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492769","url":null,"abstract":"In sphere of criminology, suspect prediction analysis has been the point of convergence for many researchers. The focus of this study is on three prime attributes of next serial suspect's biography including nationality, age and time. Generally, to prevent the uncertainty in dynamic systems by nonlinear methods, a predictor is required in Time Delay Neural Network (TDNN). However, existing TDNN with single activation function is less effective to predict labeled class due to lower accuracy. Poor approximation of smooth mapping in single hidden layer makes it less effective. This study aims to propose a combined transfer functions to improve Nonlinear Autoregressive Time Series for performance prediction with exogenous (external) input (NARX)'s by utilizing Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) algorithms. Consequently Hyperbolic Tangent Sigmoid (Tansig) and Radial Basis Function (RBF) are used in LM and SCG algorithms as bi-transfer functions for prediction of next suspect's biography in commercial serial case. The results of NARX model with combination of Tansig and RBF as two objective of transfer functions of LM and SCG, presented better performance for prediction of next serial crime suspect's biography in comparison to single activation function of Tansig and RBF.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132133402","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492813
Pavel Vlasánek, I. Perfilieva
We propose to solve the problem of image inpainting with the technique of the F-transforms, especially of the zero and first degrees. We are focused on the so called patch inpainting. The proposed technique creates a feature vector of an image characterizing weighted average levels of the corresponding intensity function and its partial derivatives over certain areas. The results of the proposed algorithm are demonstrated on various patch sizes and the issue of a reconstruction quality is discussed.
{"title":"Patch based inpainting method based on the F1-transform","authors":"Pavel Vlasánek, I. Perfilieva","doi":"10.1109/SOCPAR.2015.7492813","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492813","url":null,"abstract":"We propose to solve the problem of image inpainting with the technique of the F-transforms, especially of the zero and first degrees. We are focused on the so called patch inpainting. The proposed technique creates a feature vector of an image characterizing weighted average levels of the corresponding intensity function and its partial derivatives over certain areas. The results of the proposed algorithm are demonstrated on various patch sizes and the issue of a reconstruction quality is discussed.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131316966","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492803
Divya S. Vidyadharan, S. Thampi
Digital image tampering detection has become an active research area in the recent decade. Among the different types of image tampering, manipulation involving facial regions are of great interest as innocents are often victimized for unlawful benefits. Image splicing is a kind of image tampering where an image region is copied from one image and pasted onto another image. Principal component analysis is applied on facial regions extracted from illuminant maps to identify the facial region copy-pasted onto a group photo. Experiments were conducted on statistics-based and physics-based illumination maps and results showed that the method achieved a true positive rate of 62% and 64% respectively.
{"title":"Detecting spliced face in a group photo using PCA","authors":"Divya S. Vidyadharan, S. Thampi","doi":"10.1109/SOCPAR.2015.7492803","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492803","url":null,"abstract":"Digital image tampering detection has become an active research area in the recent decade. Among the different types of image tampering, manipulation involving facial regions are of great interest as innocents are often victimized for unlawful benefits. Image splicing is a kind of image tampering where an image region is copied from one image and pasted onto another image. Principal component analysis is applied on facial regions extracted from illuminant maps to identify the facial region copy-pasted onto a group photo. Experiments were conducted on statistics-based and physics-based illumination maps and results showed that the method achieved a true positive rate of 62% and 64% respectively.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128302786","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492795
Yuta Ueki, K. Ohnishi
Self-evaluation and evaluations among peer learners, which together are called mutual evaluations in this study, are presumed to motivate learners to increase learning and introspection. The basis behind these effects is the awareness resulting from the differences between self-evaluation and evaluations from others. Therefore, for learners to benefit from the effects of mutual evaluations, we need to easily visualize the mutual evaluation results that make them aware of something. In this paper we design a system to visualize mutual evaluation data, which are many and multi-dimensional data, in such a way that we can easily grasp the overview at first glance and develop an actual system according to the design for fostering “Fundamental Competencies for Working Persons” as an example. In addition, we compare the visualization method used in the developed system with an existing method through a subjective evaluation test and a statistical hypothesis test. Finally, we show the usefulness of our visualization method by comparing the results.
{"title":"Inducing awareness for learners through visualizing mutual evaluation data by a self-organizing map","authors":"Yuta Ueki, K. Ohnishi","doi":"10.1109/SOCPAR.2015.7492795","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492795","url":null,"abstract":"Self-evaluation and evaluations among peer learners, which together are called mutual evaluations in this study, are presumed to motivate learners to increase learning and introspection. The basis behind these effects is the awareness resulting from the differences between self-evaluation and evaluations from others. Therefore, for learners to benefit from the effects of mutual evaluations, we need to easily visualize the mutual evaluation results that make them aware of something. In this paper we design a system to visualize mutual evaluation data, which are many and multi-dimensional data, in such a way that we can easily grasp the overview at first glance and develop an actual system according to the design for fostering “Fundamental Competencies for Working Persons” as an example. In addition, we compare the visualization method used in the developed system with an existing method through a subjective evaluation test and a statistical hypothesis test. Finally, we show the usefulness of our visualization method by comparing the results.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128673011","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 : 2015-11-01DOI: 10.1109/SOCPAR.2015.7492798
Ryozo Kitajima, R. Kamimura, O. Uchida, F. Toriumi
The present paper aims to apply a new neural learning method called "Neural Potential Learning, NPL" to the classification and interpretation of tweets. It has been well known that social media such as the Twitter play crucial roles in transmitting important information at the time of natural disasters. In particular, since the Great East Japan Earthquake in 2011, the Twitter has been considered as one of the most efficient and convenient communication tools. However, because much redundant information is contained in the tweets, it is usually difficult to obtain important information from the flows of the tweets. Thus, it is urgently needed to develop some methods to extract the important and useful information from redundant tweets. To cope with complex and redundant data, a new neural potential learning has been developed to extract the important information. The method aims to find some highly potential neurons and enhance those neurons as much as possible to reduce redundant information and to focus on important information. The method was applied to the real tweets data collected in the earthquake and it was found that the method could classify the tweets as important and unimportant ones more accurately than the other conventional machine learning methods. In addition, the method made it possible to interpret how the tweets could be classified, based on the examination of highly potential neurons.
{"title":"Neural potential learning for tweets classification and interpretation","authors":"Ryozo Kitajima, R. Kamimura, O. Uchida, F. Toriumi","doi":"10.1109/SOCPAR.2015.7492798","DOIUrl":"https://doi.org/10.1109/SOCPAR.2015.7492798","url":null,"abstract":"The present paper aims to apply a new neural learning method called \"Neural Potential Learning, NPL\" to the classification and interpretation of tweets. It has been well known that social media such as the Twitter play crucial roles in transmitting important information at the time of natural disasters. In particular, since the Great East Japan Earthquake in 2011, the Twitter has been considered as one of the most efficient and convenient communication tools. However, because much redundant information is contained in the tweets, it is usually difficult to obtain important information from the flows of the tweets. Thus, it is urgently needed to develop some methods to extract the important and useful information from redundant tweets. To cope with complex and redundant data, a new neural potential learning has been developed to extract the important information. The method aims to find some highly potential neurons and enhance those neurons as much as possible to reduce redundant information and to focus on important information. The method was applied to the real tweets data collected in the earthquake and it was found that the method could classify the tweets as important and unimportant ones more accurately than the other conventional machine learning methods. In addition, the method made it possible to interpret how the tweets could be classified, based on the examination of highly potential neurons.","PeriodicalId":409493,"journal":{"name":"2015 7th International Conference of Soft Computing and Pattern Recognition (SoCPaR)","volume":"119 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2015-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129349962","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：