In tag-word disambiguation, a word is assigned to a specific context chosen among the different ones to which it is related. Relatedness to a context is often defined based on the co-occurrence of the target word with other words (context words) in sentences of a specific corpus. The overall disambiguation process can be thought as a classification process, where the context words play the role of features for the target. A problem with this approach is that the large number of possible context words can reduce the classification performance, both in terms of computational effort and in terms of quality of the outcome. Feature selection can improve the process in both regards, by reducing the overall feature space to a manageable size with high information content. In this work we propose to use, in disambiguation, a feature selection approach based on the Shapley Value (SV) - a Coalitional Game Theory related metrics, measuring the importance of a component within a coalition. By including in the feature set only the words with the highest Shapley Value, we obtain remarkable quality and performance improvements. The problem of the exponential complexity in the exact SV computation is avoided by an approximate computation based on sampling. We demonstrate the effectiveness of this method and of the sampling approach results, by using both a synthetic language corpus and a real world linguistic corpus.
{"title":"Selecting Feature-Words in Tag Sense Disambiguation Based on Their Shapley Value","authors":"Meshesha Legesse, G. Gianini, Dereje Teferi","doi":"10.1109/SITIS.2016.45","DOIUrl":"https://doi.org/10.1109/SITIS.2016.45","url":null,"abstract":"In tag-word disambiguation, a word is assigned to a specific context chosen among the different ones to which it is related. Relatedness to a context is often defined based on the co-occurrence of the target word with other words (context words) in sentences of a specific corpus. The overall disambiguation process can be thought as a classification process, where the context words play the role of features for the target. A problem with this approach is that the large number of possible context words can reduce the classification performance, both in terms of computational effort and in terms of quality of the outcome. Feature selection can improve the process in both regards, by reducing the overall feature space to a manageable size with high information content. In this work we propose to use, in disambiguation, a feature selection approach based on the Shapley Value (SV) - a Coalitional Game Theory related metrics, measuring the importance of a component within a coalition. By including in the feature set only the words with the highest Shapley Value, we obtain remarkable quality and performance improvements. The problem of the exponential complexity in the exact SV computation is avoided by an approximate computation based on sampling. We demonstrate the effectiveness of this method and of the sampling approach results, by using both a synthetic language corpus and a real world linguistic corpus.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134136904","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}
A salary prediction model was generated for graduate students using a data mining technique to generate for individuals with similar training attributes. An experiment was also conducted to compare the two data mining techniques Decision Trees ID3, C4.5 and Random Forest to determine the most suitable technique for salary prediction, tuned with key important parameters to improve the accuracy of the results. Random Forest gave the best accuracy at 90.50%, while Decision Trees ID3 and C4.5 returned lower accuracies at 61.37% and 73.96%, respectively for 13,541 records of graduate students using a 10-fold cross-validation method. Random Forest generated the best efficiency model for salary prediction. A questionnaire survey was conducted to determine usage evaluation with 50 samples. Results indicated that the system was effective in boosting students' motivation for studying, and also gave them a positive future viewpoint. The results also suggested that the students were satisfied with the implemented system since it was easy to use, and the prediction results were simple to understand without any previous background statistical knowledge.
{"title":"Random Forest for Salary Prediction System to Improve Students' Motivation","authors":"Pornthep Khongchai, Pokpong Songmuang","doi":"10.1109/SITIS.2016.106","DOIUrl":"https://doi.org/10.1109/SITIS.2016.106","url":null,"abstract":"A salary prediction model was generated for graduate students using a data mining technique to generate for individuals with similar training attributes. An experiment was also conducted to compare the two data mining techniques Decision Trees ID3, C4.5 and Random Forest to determine the most suitable technique for salary prediction, tuned with key important parameters to improve the accuracy of the results. Random Forest gave the best accuracy at 90.50%, while Decision Trees ID3 and C4.5 returned lower accuracies at 61.37% and 73.96%, respectively for 13,541 records of graduate students using a 10-fold cross-validation method. Random Forest generated the best efficiency model for salary prediction. A questionnaire survey was conducted to determine usage evaluation with 50 samples. Results indicated that the system was effective in boosting students' motivation for studying, and also gave them a positive future viewpoint. The results also suggested that the students were satisfied with the implemented system since it was easy to use, and the prediction results were simple to understand without any previous background statistical knowledge.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134429629","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}
I. D. Falco, A. D. Cioppa, U. Scafuri, E. Tarantino
In this paper the application of a general-purpose distributed Differential Evolution algorithm to range image registration is presented. The algorithm is characterized by an asynchronous migration mechanism and by a multi-population recombination information exchange, and is also supplied with adaptive updating schemes for automatically setting the Differential Evolution control parameters. In particular, this algorithm has been employed to tackle the problem of the pair-wise range image registration. Given two images with the first set as the model, the scope is to find the best possible spatial transformation of the second image allowing for 3D reconstruction of the original model. Experimental findings demonstrate the ability of such an adaptive algorithm in finding out efficient image transformations. A comparison of the results with those attained by recently presented evolutionary algorithms show the effectiveness of the proposed approach in terms of both quality and robustness of the reconstructed 3D image, and of computational cost.
{"title":"Fast Range Image Registration by an Asynchronous Adaptive Distributed Differential Evolution","authors":"I. D. Falco, A. D. Cioppa, U. Scafuri, E. Tarantino","doi":"10.1109/SITIS.2016.107","DOIUrl":"https://doi.org/10.1109/SITIS.2016.107","url":null,"abstract":"In this paper the application of a general-purpose distributed Differential Evolution algorithm to range image registration is presented. The algorithm is characterized by an asynchronous migration mechanism and by a multi-population recombination information exchange, and is also supplied with adaptive updating schemes for automatically setting the Differential Evolution control parameters. In particular, this algorithm has been employed to tackle the problem of the pair-wise range image registration. Given two images with the first set as the model, the scope is to find the best possible spatial transformation of the second image allowing for 3D reconstruction of the original model. Experimental findings demonstrate the ability of such an adaptive algorithm in finding out efficient image transformations. A comparison of the results with those attained by recently presented evolutionary algorithms show the effectiveness of the proposed approach in terms of both quality and robustness of the reconstructed 3D image, and of computational cost.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133927979","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}
Shota Takayama, Teppei Suzuki, Y. Aoki, S. Isobe, Makoto Masuda
The demand for people tracking in dense crowds is increasing, but it is a challenging problem in the computer vision field. "Crowd tracking" is extremely difficult because of hard occlusions, various motions and posture changes. In particular, we need to handle occlusions for more robust tracking. This paper discusses robust crowd tracking based on a combination of supervoxels and optical flow tracking. The SLIC based supervoxel algorithm adaptively estimates the boundary between a person and a background. Therefore, the combination of supervoxels and optical flow tracking becomes a highly reliable approach for crowd tracking. In tracking experiments, high performance is achieved for the UCF crowd dataset.
{"title":"Tracking People in Dense Crowds Using Supervoxels","authors":"Shota Takayama, Teppei Suzuki, Y. Aoki, S. Isobe, Makoto Masuda","doi":"10.1109/SITIS.2016.90","DOIUrl":"https://doi.org/10.1109/SITIS.2016.90","url":null,"abstract":"The demand for people tracking in dense crowds is increasing, but it is a challenging problem in the computer vision field. \"Crowd tracking\" is extremely difficult because of hard occlusions, various motions and posture changes. In particular, we need to handle occlusions for more robust tracking. This paper discusses robust crowd tracking based on a combination of supervoxels and optical flow tracking. The SLIC based supervoxel algorithm adaptively estimates the boundary between a person and a background. Therefore, the combination of supervoxels and optical flow tracking becomes a highly reliable approach for crowd tracking. In tracking experiments, high performance is achieved for the UCF crowd dataset.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133992132","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}
C. Stamile, F. Cotton, D. Sappey-Marinier, S. Huffel
Longitudinal analysis of neuroimaging data is becoming an important research area. In the last few years analysis of longitudinal data become a crucial point to better understand pathological mechanisms of complex brain diseases such as multiple sclerosis (MS) where white matter (WM) fiber bundles are variably altered by inflammatory events. In this work, we propose a new fully automated method to detect significant longitudinal changes in diffusivity metrics along WM fiber-bundles. This method consists of two steps: i) preprocessing of longitudinal diffusion acquisitions and WM fiber-bundles extraction, ii) application of a new hierarchical non negative matrix factorization (hNMF) algorithm to detect "pathological" changes. This method was applied first, on simulated longitudinal variations, and second, on MS patients longitudinal data. High level of precision, recall and F-Measure were obtained for the detection of small longitudinal changes along the WM fiber-bundles.
{"title":"Longitudinal Neuroimaging Analysis Using Non-Negative Matrix Factorization","authors":"C. Stamile, F. Cotton, D. Sappey-Marinier, S. Huffel","doi":"10.1109/SITIS.2016.18","DOIUrl":"https://doi.org/10.1109/SITIS.2016.18","url":null,"abstract":"Longitudinal analysis of neuroimaging data is becoming an important research area. In the last few years analysis of longitudinal data become a crucial point to better understand pathological mechanisms of complex brain diseases such as multiple sclerosis (MS) where white matter (WM) fiber bundles are variably altered by inflammatory events. In this work, we propose a new fully automated method to detect significant longitudinal changes in diffusivity metrics along WM fiber-bundles. This method consists of two steps: i) preprocessing of longitudinal diffusion acquisitions and WM fiber-bundles extraction, ii) application of a new hierarchical non negative matrix factorization (hNMF) algorithm to detect \"pathological\" changes. This method was applied first, on simulated longitudinal variations, and second, on MS patients longitudinal data. High level of precision, recall and F-Measure were obtained for the detection of small longitudinal changes along the WM fiber-bundles.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125204201","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}
Dempster-Shafer theory is a very powerful tool for data fusion, which provides a good estimation of imprecision, conflict from different sources and deal with any unions of hypotheses. In this paper, we propose to develop a high-performance hybrid Network Intrusion Detection System, based on belief functions. This system contains three levels, the first one includes two fast classifiers: Naïve Bayes and Support Vector Machine (SVM) Bused for their performance on classification. In the second level outputs of both SVM and Naïve Bayes are fuzzified using fuzzy logic. Third, the overall decision of the system is performed using Dempster's rule of combination. The experimentation on a recent benchmark dataset shows that our approach achieves a higher detection rate with low false alarm rates compared to some existing classifiers.
{"title":"On the Use of Belief Functions to Improve High Performance Intrusion Detection System","authors":"Alem Abdelkader, Y. Dahmani, A. Hadjali","doi":"10.1109/SITIS.2016.50","DOIUrl":"https://doi.org/10.1109/SITIS.2016.50","url":null,"abstract":"Dempster-Shafer theory is a very powerful tool for data fusion, which provides a good estimation of imprecision, conflict from different sources and deal with any unions of hypotheses. In this paper, we propose to develop a high-performance hybrid Network Intrusion Detection System, based on belief functions. This system contains three levels, the first one includes two fast classifiers: Naïve Bayes and Support Vector Machine (SVM) Bused for their performance on classification. In the second level outputs of both SVM and Naïve Bayes are fuzzified using fuzzy logic. Third, the overall decision of the system is performed using Dempster's rule of combination. The experimentation on a recent benchmark dataset shows that our approach achieves a higher detection rate with low false alarm rates compared to some existing classifiers.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125053812","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 paper aimed to develop a prediction model to classify problems arising in images obtained from dashboard camera video by using machine-learning algorithms. The authors generated a dataset, called the DS dataset, which contained 900 images. The dataset was divided into three groups of problems comprised of lightness problems, a combination of lightness and blur problems, and a combination of lightness and noise problems. In this study, five features on the dataset were utilised, including mean, standard deviation, entropy, histogram, and variance of the images. Classification was performed on 3 machine-learning algorithms, which were Decision Tree, Naïve Bayes and Support Vector Machines on images and partitions of the images. The experimental results showed that decision tree algorithm yielded the best performance in comparison with the two other algorithms, with the optimal prediction model obtaining an accuracy rate of up to 97.88 percent.
{"title":"Image Problem Classification for Dashboard Cameras","authors":"Narit Hnoohom, Thanchanok Thanapattherakul","doi":"10.1109/SITIS.2016.112","DOIUrl":"https://doi.org/10.1109/SITIS.2016.112","url":null,"abstract":"This paper aimed to develop a prediction model to classify problems arising in images obtained from dashboard camera video by using machine-learning algorithms. The authors generated a dataset, called the DS dataset, which contained 900 images. The dataset was divided into three groups of problems comprised of lightness problems, a combination of lightness and blur problems, and a combination of lightness and noise problems. In this study, five features on the dataset were utilised, including mean, standard deviation, entropy, histogram, and variance of the images. Classification was performed on 3 machine-learning algorithms, which were Decision Tree, Naïve Bayes and Support Vector Machines on images and partitions of the images. The experimental results showed that decision tree algorithm yielded the best performance in comparison with the two other algorithms, with the optimal prediction model obtaining an accuracy rate of up to 97.88 percent.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129250782","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}
With the development of single-person analysis in computer vision, social group analysis has received growing attention as the next area of research. In particular, group detection has been actively studied as the first step of social analysis. Here, group means an F-formation, that is, a spatial organization of people gathered for conversation. Popular group detection methods are based on coincidences in the visual attention field that are calculated from the position and body orientation of the individuals in the group. However, most previous studies have assumed that each member has the same visual attention field, and they do not consider changes in the scene over time. In this paper, we present a robust method for detection of time-varying F-formations in social space, its visual attention field model is based on the local environment. We present the results of an experiment that uses a dataset of multiple scenes, an analysis of these results validates the advantages of our method.
{"title":"Conversational Group Detection Based on Social Context Using Graph Clustering Algorithm","authors":"Shoichi Inaba, Y. Aoki","doi":"10.1109/SITIS.2016.89","DOIUrl":"https://doi.org/10.1109/SITIS.2016.89","url":null,"abstract":"With the development of single-person analysis in computer vision, social group analysis has received growing attention as the next area of research. In particular, group detection has been actively studied as the first step of social analysis. Here, group means an F-formation, that is, a spatial organization of people gathered for conversation. Popular group detection methods are based on coincidences in the visual attention field that are calculated from the position and body orientation of the individuals in the group. However, most previous studies have assumed that each member has the same visual attention field, and they do not consider changes in the scene over time. In this paper, we present a robust method for detection of time-varying F-formations in social space, its visual attention field model is based on the local environment. We present the results of an experiment that uses a dataset of multiple scenes, an analysis of these results validates the advantages of our method.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129476144","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}
Identifying different types of damage is very essential in times of natural disasters, where first responders are flooding the internet with often annotated images and texts, and rescue teams are overwhelmed to prioritize often scarce resources. While most of the efforts in such humanitarian situations rely heavily on human labor and input, we propose in this paper a novel hybrid approach to help automate more humanitarian computing. Our framework merges low-level visual features that extract color, shape and texture along with a semantic attribute that is obtained after comparing the picture annotation to some bag of words. These visual and textual features were trained and tested on a dataset gathered from the SUN database and some Google Images. The best accuracy obtained using low-level features alone is 91.3 %, while appending the semantic attributes to it raised the accuracy to 95.5% using linear SVM and 5-Fold cross-validation which motivates an updated folk statement "an ANNOTATED image is worth a thousand word ".
{"title":"Semantic and Visual Cues for Humanitarian Computing of Natural Disaster Damage Images","authors":"H. Jomaa, Yara Rizk, M. Awad","doi":"10.1109/SITIS.2016.70","DOIUrl":"https://doi.org/10.1109/SITIS.2016.70","url":null,"abstract":"Identifying different types of damage is very essential in times of natural disasters, where first responders are flooding the internet with often annotated images and texts, and rescue teams are overwhelmed to prioritize often scarce resources. While most of the efforts in such humanitarian situations rely heavily on human labor and input, we propose in this paper a novel hybrid approach to help automate more humanitarian computing. Our framework merges low-level visual features that extract color, shape and texture along with a semantic attribute that is obtained after comparing the picture annotation to some bag of words. These visual and textual features were trained and tested on a dataset gathered from the SUN database and some Google Images. The best accuracy obtained using low-level features alone is 91.3 %, while appending the semantic attributes to it raised the accuracy to 95.5% using linear SVM and 5-Fold cross-validation which motivates an updated folk statement \"an ANNOTATED image is worth a thousand word \".","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"106 4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120994865","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}
We present a hierarchical self-organizing map based system for online recognition of human actions. We have made a first evaluation of our system by training it on two different sets of recorded human actions, one set containing manner actions and one set containing result actions, and then tested it by letting a human performer carry out the actions online in real time in front of the system's 3D-camera. The system successfully recognized more than 94% of the manner actions and most of the result actions carried out by the human performer.
{"title":"Action Recognition Online with Hierarchical Self-Organizing Maps","authors":"Zahra Gharaee, P. Gärdenfors, Magnus Johnsson","doi":"10.1109/SITIS.2016.91","DOIUrl":"https://doi.org/10.1109/SITIS.2016.91","url":null,"abstract":"We present a hierarchical self-organizing map based system for online recognition of human actions. We have made a first evaluation of our system by training it on two different sets of recorded human actions, one set containing manner actions and one set containing result actions, and then tested it by letting a human performer carry out the actions online in real time in front of the system's 3D-camera. The system successfully recognized more than 94% of the manner actions and most of the result actions carried out by the human performer.","PeriodicalId":403704,"journal":{"name":"2016 12th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114269031","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}