Pub Date : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00049
Daniyal Kazempour, Peer Kröger, T. Seidl
In the setting of unsupervised machine learning, especially in clustering tasks, the evaluation of either novel algorithms or the assessment of a clustering of novel data is challenging. While mostly in the literature the evaluation of new methods is performed on labelled data, there are cases where no labels are at our disposal. In other cases we may not want to trust the “ground truth” labels. In general there exists a spectrum of so called internal evaluation measures in the literature. Each of the measures is mostly specialized towards a specific clustering model. The model of arbitrarily oriented subspace clusters is a more recent one. To the best of our knowledge there exist at the current time no internal evaluation measures tailored at assessing this particular type of clusterings. In this work we present the first internal quality measures for arbitrarily oriented subspace clusterings namely the normalized projected energy (NPE) and subspace compactness score (SCS). The results from the experiments show that especially NPE is capable of assessing clusterings by considering archetypical properties of arbitrarily oriented subspace clustering.
{"title":"Towards an Internal Evaluation Measure for Arbitrarily Oriented Subspace Clustering","authors":"Daniyal Kazempour, Peer Kröger, T. Seidl","doi":"10.1109/ICDMW51313.2020.00049","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00049","url":null,"abstract":"In the setting of unsupervised machine learning, especially in clustering tasks, the evaluation of either novel algorithms or the assessment of a clustering of novel data is challenging. While mostly in the literature the evaluation of new methods is performed on labelled data, there are cases where no labels are at our disposal. In other cases we may not want to trust the “ground truth” labels. In general there exists a spectrum of so called internal evaluation measures in the literature. Each of the measures is mostly specialized towards a specific clustering model. The model of arbitrarily oriented subspace clusters is a more recent one. To the best of our knowledge there exist at the current time no internal evaluation measures tailored at assessing this particular type of clusterings. In this work we present the first internal quality measures for arbitrarily oriented subspace clusterings namely the normalized projected energy (NPE) and subspace compactness score (SCS). The results from the experiments show that especially NPE is capable of assessing clusterings by considering archetypical properties of arbitrarily oriented subspace clustering.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125296323","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00034
Tomas Sousa Pereira, T. Cunha, C. Soares
Collaborative Filtering (CF) has become the standard approach to solve recommendation systems problems. Collaborative Filtering algorithms try to make predictions about interests of a user by collecting the personal interests from multiple users. There are multiple CF algorithms, each one of them with its own biases. It is the Machine Learning practitioner that has to choose the best algorithm for each task beforehand. In Recommender Systems, different algorithms have different performance for different users within the same dataset. Meta Learning has been used to choose the best algorithm for a given problem. Meta Learning is usually applied to select algorithms for a whole dataset. Adapting it to select the to the algorithm for a single user in a RS involves several challenges. The most important is the design of the metafeatures which, in typical meta learning, characterize datasets while here, they must characterize a single user. This work presents a new meta-learning based framework named $mu-mathbf{cf}2mathbf{vec}$ to select the best algorithm for each user. We propose using Representation Learning techniques to extract the metafeatures. Representation Learning tries to extract representations that can be reused in other learning tasks. In this work we also implement the framework using different RL techniques to evaluate which one can be more useful to solve this task. In the meta level, the meta learning model will use the metafeatures to extract knowledge that will be used to predict the best algorithm for each user. We evaluated an implementation of this framework using MovieLens 20M dataset. Our implementation achieved consistent gains in the meta level, however, in the base level we only achieved marginal gains.
{"title":"$mu-text{cf}2text{vec}$: Representation Learning for Personalized Algorithm Selection in Recommender Systems","authors":"Tomas Sousa Pereira, T. Cunha, C. Soares","doi":"10.1109/ICDMW51313.2020.00034","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00034","url":null,"abstract":"Collaborative Filtering (CF) has become the standard approach to solve recommendation systems problems. Collaborative Filtering algorithms try to make predictions about interests of a user by collecting the personal interests from multiple users. There are multiple CF algorithms, each one of them with its own biases. It is the Machine Learning practitioner that has to choose the best algorithm for each task beforehand. In Recommender Systems, different algorithms have different performance for different users within the same dataset. Meta Learning has been used to choose the best algorithm for a given problem. Meta Learning is usually applied to select algorithms for a whole dataset. Adapting it to select the to the algorithm for a single user in a RS involves several challenges. The most important is the design of the metafeatures which, in typical meta learning, characterize datasets while here, they must characterize a single user. This work presents a new meta-learning based framework named $mu-mathbf{cf}2mathbf{vec}$ to select the best algorithm for each user. We propose using Representation Learning techniques to extract the metafeatures. Representation Learning tries to extract representations that can be reused in other learning tasks. In this work we also implement the framework using different RL techniques to evaluate which one can be more useful to solve this task. In the meta level, the meta learning model will use the metafeatures to extract knowledge that will be used to predict the best algorithm for each user. We evaluated an implementation of this framework using MovieLens 20M dataset. Our implementation achieved consistent gains in the meta level, however, in the base level we only achieved marginal gains.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125390671","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00112
Elliott Ash, J. Jacobs, Bentley Macleod, S. Naidu, Dominik Stammbach
This paper describes an unsupervised legal document parser which performs a decomposition of labor union contracts into discrete assignments of rights and duties among agents of interest. We use insights from deontic logic applied to modal categories and other linguistic patterns to generate topic-specific measures of relative legal authority. We illustrate the consistency and efficiency of the pipeline by applying it to a large corpus of 35K contracts and validating the resulting outputs.
{"title":"Unsupervised Extraction of Workplace Rights and Duties from Collective Bargaining Agreements","authors":"Elliott Ash, J. Jacobs, Bentley Macleod, S. Naidu, Dominik Stammbach","doi":"10.1109/ICDMW51313.2020.00112","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00112","url":null,"abstract":"This paper describes an unsupervised legal document parser which performs a decomposition of labor union contracts into discrete assignments of rights and duties among agents of interest. We use insights from deontic logic applied to modal categories and other linguistic patterns to generate topic-specific measures of relative legal authority. We illustrate the consistency and efficiency of the pipeline by applying it to a large corpus of 35K contracts and validating the resulting outputs.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122715356","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00037
Ahmad Shahzad, Frans Coenen
This paper presents the Distributed Kruskal Algorithm for Minimum Spanning Tree (MST) based clustering to be used in the context of recommendation engines. The algorithm can operate over large graph data sets distributed over a number of machines. The operation of the algorithm is evaluated by comparing both the quality of the cluster configurations produced, and the accuracy of the predictions, with non-MST based clustering approaches. The results indicate that the proposed approach produces comparable recommendations at much lower storage, hence runtime, costs.
{"title":"Efficient Distributed MST Based Clustering for Recommender Systems","authors":"Ahmad Shahzad, Frans Coenen","doi":"10.1109/ICDMW51313.2020.00037","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00037","url":null,"abstract":"This paper presents the Distributed Kruskal Algorithm for Minimum Spanning Tree (MST) based clustering to be used in the context of recommendation engines. The algorithm can operate over large graph data sets distributed over a number of machines. The operation of the algorithm is evaluated by comparing both the quality of the cluster configurations produced, and the accuracy of the predictions, with non-MST based clustering approaches. The results indicate that the proposed approach produces comparable recommendations at much lower storage, hence runtime, costs.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129604326","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00133
M. Marinescu, Artem Reshetnikov, J. M. López
This paper proposes a novel approach to object detection for the Cultural Heritage domain, which relies on combining Deep Learning and semantic metadata about candidate objects extracted from existing sources such as Wikidata, dictionaries, or Google NGram. Working with cultural heritage presents challenges not present in every-day images. In computer vision, object detection models are usually trained with datasets whose classes are not imaginary concepts, and have neither symbolic nor time-specific dimensions. Apart from this conceptual problem, the paintings are limited in number and represent the same concept in potentially very different styles. Finally, the metadata associated with the images is often poor or inexistent, which makes it hard to properly train a model. Our approach can improve the precision of object detection by placing the classes detected by a neural network model in time, based on the dates of their first known use. By taking into account the time of inception of objects such as the TV, cell phone, or scissors, and the appearance of some objects in the geographical space that corresponds to a painting (e.g. bananas or broccoli in 15th century Europe), we can correct and refine the detected objects based on their chronologic probability.
{"title":"Improving object detection in paintings based on time contexts","authors":"M. Marinescu, Artem Reshetnikov, J. M. López","doi":"10.1109/ICDMW51313.2020.00133","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00133","url":null,"abstract":"This paper proposes a novel approach to object detection for the Cultural Heritage domain, which relies on combining Deep Learning and semantic metadata about candidate objects extracted from existing sources such as Wikidata, dictionaries, or Google NGram. Working with cultural heritage presents challenges not present in every-day images. In computer vision, object detection models are usually trained with datasets whose classes are not imaginary concepts, and have neither symbolic nor time-specific dimensions. Apart from this conceptual problem, the paintings are limited in number and represent the same concept in potentially very different styles. Finally, the metadata associated with the images is often poor or inexistent, which makes it hard to properly train a model. Our approach can improve the precision of object detection by placing the classes detected by a neural network model in time, based on the dates of their first known use. By taking into account the time of inception of objects such as the TV, cell phone, or scissors, and the appearance of some objects in the geographical space that corresponds to a painting (e.g. bananas or broccoli in 15th century Europe), we can correct and refine the detected objects based on their chronologic probability.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124716779","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00100
Hyunjin Kim, Dongseop Lee, Jaecheol Ryou
In a smart energy environment, user authentication is an essential process in smart energy. After the authentication process to verify whether the user is registered one, the user can access the smart service such as power consumption prediction, intelligent energy management. The user authentication technology has evolved to various authentication methods using ID/PW, security token and biometric information because of the diversification of the Internet of Things and social structure. Despite there are various authentication technologies, the ID/PW authentication still widely used because of low cost and convenience. However, the user using ID/PW methods should use different passwords for each service, as well as passwords that include special symbols. Moreover, it is difficult for users to remember complicated passwords and it is not easy to change passwords periodically. Therefore, in this paper, we propose the user authentication method using FIDO based password management. Through the password management, the user login the information system by password as well as biometric information using hardware security device. In addition, the method is compatible with legacy PW authentication. The proposed mechanism will enhance the security on ID/PW authentication method currently in use on most service.
{"title":"User Authentication Method using FIDO based Password Management for Smart Energy Environment","authors":"Hyunjin Kim, Dongseop Lee, Jaecheol Ryou","doi":"10.1109/ICDMW51313.2020.00100","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00100","url":null,"abstract":"In a smart energy environment, user authentication is an essential process in smart energy. After the authentication process to verify whether the user is registered one, the user can access the smart service such as power consumption prediction, intelligent energy management. The user authentication technology has evolved to various authentication methods using ID/PW, security token and biometric information because of the diversification of the Internet of Things and social structure. Despite there are various authentication technologies, the ID/PW authentication still widely used because of low cost and convenience. However, the user using ID/PW methods should use different passwords for each service, as well as passwords that include special symbols. Moreover, it is difficult for users to remember complicated passwords and it is not easy to change passwords periodically. Therefore, in this paper, we propose the user authentication method using FIDO based password management. Through the password management, the user login the information system by password as well as biometric information using hardware security device. In addition, the method is compatible with legacy PW authentication. The proposed mechanism will enhance the security on ID/PW authentication method currently in use on most service.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130532392","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00070
Clinton Daniel, T. Gill, A. Hevner, Matthew T. Mullarkey
Security Analysts (SAs) operating within Security Operation Centers (SOCs) conduct cybersecurity investigations on cyber events using methods which pave a measurable path. These paths serve as a source of evidence to study the transitions of the cognitive tasks performed by the SA throughout the investigation. Insight into these paths can support the observation and understanding of how to evaluate and measure the critical decisions made during an investigation such as when a SA transitions from analyzing event logs to observing threat intelligence. We propose a framework we call the Cyber Analysis Transition Framework which applies a quantitative approach for evaluating and measuring the transitions of the SA conducting cyber analysis methods. The novel approach for this framework includes the application of process mining and deep neural network output as a means for evaluating and measuring a SA's performance while conducting cybersecurity investigations.
{"title":"A Deep Neural Network Approach to Tracing Paths in Cybersecurity Investigations","authors":"Clinton Daniel, T. Gill, A. Hevner, Matthew T. Mullarkey","doi":"10.1109/ICDMW51313.2020.00070","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00070","url":null,"abstract":"Security Analysts (SAs) operating within Security Operation Centers (SOCs) conduct cybersecurity investigations on cyber events using methods which pave a measurable path. These paths serve as a source of evidence to study the transitions of the cognitive tasks performed by the SA throughout the investigation. Insight into these paths can support the observation and understanding of how to evaluate and measure the critical decisions made during an investigation such as when a SA transitions from analyzing event logs to observing threat intelligence. We propose a framework we call the Cyber Analysis Transition Framework which applies a quantitative approach for evaluating and measuring the transitions of the SA conducting cyber analysis methods. The novel approach for this framework includes the application of process mining and deep neural network output as a means for evaluating and measuring a SA's performance while conducting cybersecurity investigations.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124028476","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00025
K. Yada, Ken Ishibashi, Taku Ohashi, Danhua Wang, S. Tsumoto
The purpose of this study was to classify shopping trip types based on customer path data and to identify differences in effectiveness of sales promotions. Existing studies on shopping trip types have not incorporated customer in-store behavior data as an index for classification. In this paper, we categorize customer shopping trip types into two categories of “major trip” and “fill-in trip”, and investigate the differences in the impact of sales promotions on sales effectiveness using customer path data. Impact of sales is measured by the probability of occurrence in the three processes of the purchase process, based on existing research.
{"title":"How Shoppers Walk and Shop in a Supermarket","authors":"K. Yada, Ken Ishibashi, Taku Ohashi, Danhua Wang, S. Tsumoto","doi":"10.1109/ICDMW51313.2020.00025","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00025","url":null,"abstract":"The purpose of this study was to classify shopping trip types based on customer path data and to identify differences in effectiveness of sales promotions. Existing studies on shopping trip types have not incorporated customer in-store behavior data as an index for classification. In this paper, we categorize customer shopping trip types into two categories of “major trip” and “fill-in trip”, and investigate the differences in the impact of sales promotions on sales effectiveness using customer path data. Impact of sales is measured by the probability of occurrence in the three processes of the purchase process, based on existing research.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"42 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127592609","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 : 2020-11-01DOI: 10.1109/ICDMW51313.2020.00040
Hangbin Zhang, R. Wong, Victor W. Chu
Latent variable models have been widely adopted by recommender systems due to the advancements of their learning scalability and performance. Recent research has focused on hybrid models. However, due to the sparsity of user and/or item data, most of these proposals have convoluted model architectures and objective functions. In particular, the latter are mostly tailored for sparse data from either user or item spaces. Although it is possible to derive an analogous model for both spaces, this makes a system overly complicated. To address this problem, we propose a deep learning based latent model called Distilled Hybrid Network (DHN) with a teacher-student learning architecture. Unlike other related work that tried to better incorporate content components to improve accuracy, we instead focus on model learning optimization. To the best of our knowledge, we are the first to employ teacher-student learning architecture for recommender systems. Experiment results show that our proposed model notably outperforms state-of-the-art approaches. We also show that our proposed architecture can be applied to existing recommender models to improve their accuracies.
{"title":"Hybrid Learning with Teacher-student Knowledge Distillation for Recommenders","authors":"Hangbin Zhang, R. Wong, Victor W. Chu","doi":"10.1109/ICDMW51313.2020.00040","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00040","url":null,"abstract":"Latent variable models have been widely adopted by recommender systems due to the advancements of their learning scalability and performance. Recent research has focused on hybrid models. However, due to the sparsity of user and/or item data, most of these proposals have convoluted model architectures and objective functions. In particular, the latter are mostly tailored for sparse data from either user or item spaces. Although it is possible to derive an analogous model for both spaces, this makes a system overly complicated. To address this problem, we propose a deep learning based latent model called Distilled Hybrid Network (DHN) with a teacher-student learning architecture. Unlike other related work that tried to better incorporate content components to improve accuracy, we instead focus on model learning optimization. To the best of our knowledge, we are the first to employ teacher-student learning architecture for recommender systems. Experiment results show that our proposed model notably outperforms state-of-the-art approaches. We also show that our proposed architecture can be applied to existing recommender models to improve their accuracies.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126267941","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}
Recently, many joint deep clustering methods, which simultaneously learn latent embedding and predict clustering assignments through deep neural network, have received a lot of attention. Among these methods, KL divergence based clustering framework is one of the most popular branches. However, the clustering performances of these methods depend on an additional auxiliary target distribution. In this paper, we build a novel deep fuzzy clustering (DFC) network to learn discriminative and balanced assignment without the need of any auxiliary distribution. Specifically, we design an elaborate fuzzy clustering layer (FCL) to estimate more discriminative assignments, and utilize weighted intra-class variance (WIV) as clustering objective function to enhance the compactness of the learned embedding. Moreover, we propose extended mutual information (EMI) between input data and the corresponding clustering assignments as a regularization to achieve “fair” but “firm” assignment. Extensive experiments conducted on several datasets illustrate the superiority of the proposed approach comparing to the state-of-the-art methods.
{"title":"Deep Fuzzy Clustering with Weighted Intra-class Variance and Extended Mutual Information Regularization","authors":"Yunsheng Pang, Feiyu Chen, Sheng Huang, Yongxin Ge, Wei Wang, Taiping Zhang","doi":"10.1109/ICDMW51313.2020.00137","DOIUrl":"https://doi.org/10.1109/ICDMW51313.2020.00137","url":null,"abstract":"Recently, many joint deep clustering methods, which simultaneously learn latent embedding and predict clustering assignments through deep neural network, have received a lot of attention. Among these methods, KL divergence based clustering framework is one of the most popular branches. However, the clustering performances of these methods depend on an additional auxiliary target distribution. In this paper, we build a novel deep fuzzy clustering (DFC) network to learn discriminative and balanced assignment without the need of any auxiliary distribution. Specifically, we design an elaborate fuzzy clustering layer (FCL) to estimate more discriminative assignments, and utilize weighted intra-class variance (WIV) as clustering objective function to enhance the compactness of the learned embedding. Moreover, we propose extended mutual information (EMI) between input data and the corresponding clustering assignments as a regularization to achieve “fair” but “firm” assignment. Extensive experiments conducted on several datasets illustrate the superiority of the proposed approach comparing to the state-of-the-art methods.","PeriodicalId":426846,"journal":{"name":"2020 International Conference on Data Mining Workshops (ICDMW)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131634622","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：