The Experts below are selected from a list of 138 Experts worldwide ranked by ideXlab platform

Hong Wenxue - One of the best experts on this subject based on the ideXlab platform.

  • FSKD (1) - A Novel Visual Combining Classifier Based on a Two-dimensional Graphical Representation of the Attribute Data
    2009 Sixth International Conference on Fuzzy Systems and Knowledge Discovery, 2009
    Co-Authors: Hong Wenxue
    Abstract:

    A novel Visual Combining Classifier (VCC), which integrates two-dimensional graphical representation of the attribute data, image processing and pattern recognition techniques together, has been proposed. The basic principle of the VCC is mapping attribute data of a data matrix to the two-dimensional graphs, transforming these graphs to sub Classifiers by pixel graphs, and Combining the sub Classifiers by decision rules. By interactive approaches, the optimum graphs for classification could be chosen and then pattern recognition could be realized automatically. The two experiments of the scatter and pole graphical representations based on Iris database have been made and classification precisions are 98.67% and 97.33% by LOOCV respectively.

  • A Novel Visual Combining Classifier Based on a Two-dimensional Graphical Representation of the Attribute Data
    2009 Sixth International Conference on Fuzzy Systems and Knowledge Discovery, 2009
    Co-Authors: Hong Wenxue
    Abstract:

    A novel visual Combining Classifier (VCC), which integrates two-dimensional graphical representation of the attribute data, image processing and pattern recognition techniques together, has been proposed. The basic principle of the VCC is mapping attribute data of a data matrix to the two-dimensional graphs, transforming these graphs to sub Classifiers by pixel graphs, and Combining the sub Classifiers by decision rules. By interactive approaches, the optimum graphs for classification could be chosen and then pattern recognition could be realized automatically. The two experiments of the scatter and pole graphical representations based on Iris database have been made and classification precisions are 98.67% and 97.33% by LOOCV respectively.

Marcin Zmyslony - One of the best experts on this subject based on the ideXlab platform.

  • ACIIDS (2) - Combining Classifier with a fuser implemented as a one layer perceptron
    Intelligent Information and Database Systems, 2011
    Co-Authors: Michal Wozniak, Marcin Zmyslony
    Abstract:

    The Combining approach to classification so-called Multiple Classifier Systems (MCSs) is nowadays one of the most promising directions in pattern recognition and gained a lot of interest through recent years. A large variety of methods that exploit the strengths of individual Classifiers have been developed. The most popular methods have their origins in voting, where the decision of a common Classifier is a combination of individual Classifiers' outputs, i.e. class numbers or values of discriminants. Of course to improve performance and robustness of compound Classifiers, different and diverse individual Classifiers should be combined. This work focuses on the problem of fuser design. We present some new results of our research and propose to train a fusion block by algorithms that have their origin in neural computing. As we have shown in previous works, we can produce better results Combining Classifiers than by using the abstract model of fusion so-called Oracle. The results of our experiments are presented to confirm our previous observations.

  • designing Combining Classifier with trained fuser analytical and experimental evaluation
    Intelligent Systems Design and Applications, 2010
    Co-Authors: Michal Wozniak, Marcin Zmyslony
    Abstract:

    Combining pattern recognition is the promising direction in designing an effective Classifier systems. There are several approaches of collective decision-making, among them voting methods, where the decision is a combination of individual Classifiers' outputs are quite popular. This article focuses on the problem of fuser design which uses continuous outputs of individual Classifiers to make a decision. We formulate problem of fuser design as an optimization task and use neural approach as its solver. We propose a taxonomy of aforementioned fusers and their main features are presented for some of them. The results of computer experiments carried out on benchmark datasets confirm quality of proposed concept.

  • Designing Combining Classifier with trained fuser — Analytical and experimental evaluation
    2010 10th International Conference on Intelligent Systems Design and Applications, 2010
    Co-Authors: Michal Wozniak, Marcin Zmyslony
    Abstract:

    Combining pattern recognition is the promising direction in designing an effective Classifier systems. There are several approaches of collective decision-making, among them voting methods, where the decision is a combination of individual Classifiers' outputs are quite popular. This article focuses on the problem of fuser design which uses continuous outputs of individual Classifiers to make a decision. We formulate problem of fuser design as an optimization task and use neural approach as its solver. We propose a taxonomy of aforementioned fusers and their main features are presented for some of them. The results of computer experiments carried out on benchmark datasets confirm quality of proposed concept.

Michal Wozniak - One of the best experts on this subject based on the ideXlab platform.

  • ACIIDS (2) - Combining Classifier with a fuser implemented as a one layer perceptron
    Intelligent Information and Database Systems, 2011
    Co-Authors: Michal Wozniak, Marcin Zmyslony
    Abstract:

    The Combining approach to classification so-called Multiple Classifier Systems (MCSs) is nowadays one of the most promising directions in pattern recognition and gained a lot of interest through recent years. A large variety of methods that exploit the strengths of individual Classifiers have been developed. The most popular methods have their origins in voting, where the decision of a common Classifier is a combination of individual Classifiers' outputs, i.e. class numbers or values of discriminants. Of course to improve performance and robustness of compound Classifiers, different and diverse individual Classifiers should be combined. This work focuses on the problem of fuser design. We present some new results of our research and propose to train a fusion block by algorithms that have their origin in neural computing. As we have shown in previous works, we can produce better results Combining Classifiers than by using the abstract model of fusion so-called Oracle. The results of our experiments are presented to confirm our previous observations.

  • designing Combining Classifier with trained fuser analytical and experimental evaluation
    Intelligent Systems Design and Applications, 2010
    Co-Authors: Michal Wozniak, Marcin Zmyslony
    Abstract:

    Combining pattern recognition is the promising direction in designing an effective Classifier systems. There are several approaches of collective decision-making, among them voting methods, where the decision is a combination of individual Classifiers' outputs are quite popular. This article focuses on the problem of fuser design which uses continuous outputs of individual Classifiers to make a decision. We formulate problem of fuser design as an optimization task and use neural approach as its solver. We propose a taxonomy of aforementioned fusers and their main features are presented for some of them. The results of computer experiments carried out on benchmark datasets confirm quality of proposed concept.

  • Designing Combining Classifier with trained fuser — Analytical and experimental evaluation
    2010 10th International Conference on Intelligent Systems Design and Applications, 2010
    Co-Authors: Michal Wozniak, Marcin Zmyslony
    Abstract:

    Combining pattern recognition is the promising direction in designing an effective Classifier systems. There are several approaches of collective decision-making, among them voting methods, where the decision is a combination of individual Classifiers' outputs are quite popular. This article focuses on the problem of fuser design which uses continuous outputs of individual Classifiers to make a decision. We formulate problem of fuser design as an optimization task and use neural approach as its solver. We propose a taxonomy of aforementioned fusers and their main features are presented for some of them. The results of computer experiments carried out on benchmark datasets confirm quality of proposed concept.

  • Untrained Multiple Classifier System for designing security scanner for web applications
    2010 Fifth International Conference on Information and Automation for Sustainability, 2010
    Co-Authors: Szymon Sztajer, Michal Wozniak
    Abstract:

    The Multiple Classifier Systems are nowadays one of the most promising directions in pattern recognition. There are many methods of decision making based on Classifier groups. The most popular are those methods that have their origin in voting, where the decision of the common Classifier is a combination of simple Classifiers decisions. The paper presents an idea how a decision about attack in application layer could be made by an Intrusion Detection System application using a Combining Classifier. The results of the computer experiments carried out on computer-generated data, confirm quality of the proposed concept.

Baoliang Lu - One of the best experts on this subject based on the ideXlab platform.

  • analysis of fault tolerance of a Combining Classifier
    International Symposium on Neural Networks, 2004
    Co-Authors: Hai Zhao, Baoliang Lu
    Abstract:

    This paper mainly analyses the fault tolerant capability of a Combining Classifier that uses a K-voting strategy for integrating binary Classifiers. From the point view of fault tolerance, we discuss the influence of the failure of binary Classifiers on the final output of the Combining Classifier, and present a theoretical analysis of combination performance under three fault models. The results provide a theoretical base for fault detection of the Combining Classifier.

  • ISNN (1) - Analysis of Fault Tolerance of a Combining Classifier
    Advances in Neural Networks – ISNN 2004, 2004
    Co-Authors: Hai Zhao, Baoliang Lu
    Abstract:

    This paper mainly analyses the fault tolerant capability of a Combining Classifier that uses a K-voting strategy for integrating binary Classifiers. From the point view of fault tolerance, we discuss the influence of the failure of binary Classifiers on the final output of the Combining Classifier, and present a theoretical analysis of combination performance under three fault models. The results provide a theoretical base for fault detection of the Combining Classifier.

Shrawan Kumar Trivedi - One of the best experts on this subject based on the ideXlab platform.

  • A Combining Classifiers Approach for Detecting Email Spams
    2016 30th International Conference on Advanced Information Networking and Applications Workshops (WAINA), 2016
    Co-Authors: Shrawan Kumar Trivedi
    Abstract:

    Email is a rapid and cheap communication medium for sending and receiving information where spam is becoming a nuisance for such communication. A good spam filtering cannot only be achieved by high performance accuracy but low false positive is also necessary. This paper presents a Combining Classifiers approach with committee selection mechanism where the main objective is to combine individual decisions of the good Classifiers for utmost classification outcome in spam classification domain. In this context, three different Classifiers have been selected i.e. "Boosted Bayesian", "Boosted Naïve Bayes and Support Vector Machine (SVM). For Combining Classifiers, boosted bayesian and boosted naïve bayes are chosen as members of committee and SVM is taken as the president. The member of committee have been selected from our previous study where we have identified boosting with adaboost improves the performance of probabilistic Classifier. Results show the best results of novel Combining Classifier approach in compression with individual Classifiers compared in terms of good performance accuracy and low false positives. In addition, greedy step wise feature search method is found to be good in this study.

  • AINA Workshops - A Combining Classifiers Approach for Detecting Email Spams
    2016 30th International Conference on Advanced Information Networking and Applications Workshops (WAINA), 2016
    Co-Authors: Shrawan Kumar Trivedi
    Abstract:

    Email is a rapid and cheap communication medium for sending and receiving information where spam is becoming a nuisance for such communication. A good spam filtering cannot only be achieved by high performance accuracy but low false positive is also necessary. This paper presents a Combining Classifiers approach with committee selection mechanism where the main objective is to combine individual decisions of the good Classifiers for utmost classification outcome in spam classification domain. In this context, three different Classifiers have been selected i.e. "Boosted Bayesian", "Boosted Naive Bayes and Support Vector Machine (SVM). For Combining Classifiers, boosted bayesian and boosted naive bayes are chosen as members of committee and SVM is taken as the president. The member of committee have been selected from our previous study where we have identified boosting with adaboost improves the performance of probabilistic Classifier. Results show the best results of novel Combining Classifier approach in compression with individual Classifiers compared in terms of good performance accuracy and low false positives. In addition, greedy step wise feature search method is found to be good in this study.