Showing papers by "Xizhao Wang published in 2017"
TL;DR: A novel fuzziness based semi-supervised learning approach by utilizing unlabeled samples assisted with supervised learning algorithm to improve the classifier's performance for the IDSs is proposed.
460 citations
TL;DR: Two diversity criteria, i.e., clustering-based diversity and fuzzy rough set based diversity, are proposed for MIAL by utilizing a support vector machine (SVM) based MIL classifier and the lower approximations in fuzzy rough sets are used to define a new concept named dissimilarity degree.
Abstract: Multiple-instance active learning (MIAL) is a paradigm to collect sufficient training bags for a multiple-instance learning (MIL) problem, by selecting and querying the most valuable unlabeled bags iteratively. Existing works on MIAL evaluate an unlabeled bag by its informativeness with regard to the current classifier, but neglect the internal distribution of its instances, which can reflect the diversity of the bag. In this paper, two diversity criteria, i.e., clustering-based diversity and fuzzy rough set based diversity, are proposed for MIAL by utilizing a support vector machine (SVM) based MIL classifier. In the first criterion, a kernel $k$ -means clustering algorithm is used to explore the hidden structure of the instances in the feature space of the SVM, and the diversity degree of an unlabeled bag is measured by the number of unique clusters covered by the bag. In the second criterion, the lower approximations in fuzzy rough sets are used to define a new concept named dissimilarity degree, which depicts the uniqueness of an instance so as to measure the diversity degree of a bag. By incorporating the proposed diversity criteria with existing informativeness measurements, new MIAL algorithms are developed, which can select bags with both high informativeness and diversity. Experimental comparisons demonstrate the feasibility and effectiveness of the proposed methods.
97 citations
TL;DR: It is concluded that Li's reduction maintaining the non-redundant decision rules of a formal decision context is coarser than others, and the results obtained are beneficial for users to select an appropriate reduction method for meeting their requirements.
82 citations
TL;DR: A ranking-based adaptive ABC algorithm (ARABC), in which food sources are selected by bees to search, and the parent food sources used in the solution search equation are all chosen based on their rankings, which is adaptively adjusted according to the status of the population evolution.
79 citations
TL;DR: The effectiveness of the new method based on information entropy in fuzzy incomplete information system is verified by comparing the average fusion method and an illustrative example is delivered to illustrate the effectiveness of this proposed fusion method.
Abstract: With the development of society, although the way that people get information more and more convenient, the information which people get may be incomplete and has a little degree of uncertainty and fuzziness. In real life, the incomplete fuzzy phenomenon of information source exists widely. It is extremely meaningful to fuse multiple fuzzy incomplete information sources effectively. In this study, a new method is presented for information fusion based on information entropy in fuzzy incomplete information system and the effectiveness of the new method is verified by comparing the average fusion method. Then, an illustrative example is delivered to illustrate the effectiveness of the proposed fusion method. Finally, we have also tested the veracity and validity of this method by experiment on a dataset from UCI. The results of this study will be useful for pooling the uncertain data from different information sources and significant for establishing a distinct direction of the fusion method.
55 citations
TL;DR: This work proposes a generalization of ELM for processing the monotonic classification, named as Monotonic Classification Extreme Learning Machine (MCELM) in which the monotonicity constraints are imposed to the original ELM model.
39 citations
TL;DR: The paper confirms the feasibility and effectiveness of designing discrete differential evolution algorithms for D{0-1}KP by encoding conversion approaches.
Abstract: This paper first proposes a discrete differential evolution algorithm for discounted {0-1} knapsack problems (D{0-1}KP) based on feasible solutions represented by the 0-1 vector. Subsequently, based on two encoding mechanisms of transforming a real vector into an integer vector, two new algorithms for solving D{0-1}KP are given through using integer vectors defined on {0, 1, 2, 3}n to represent feasible solutions of the problem. Finally, the paper conducts a comparative study on the performance between our proposed three discrete differential evolution algorithms and those developed by common genetic algorithms for solving several types of large scale D{0-1}KP problems. The paper confirms the feasibility and effectiveness of designing discrete differential evolution algorithms for D{0-1}KP by encoding conversion approaches.
37 citations
TL;DR: Experimental results demonstrate that the proposed selective ensemble method is superior to the single SVDD and the other four related ensemble approaches.
15 citations
TL;DR: Theoretical analysis and experimental demonstration show that the semi-supervised learning model introduced can effectively improve the performance of a cost-sensitive learning algorithm for a certain type of classifiers.
15 citations
TL;DR: This study presents a novel algorithm for categorizing the instances into three groups that can effectively integrate with D&C strategy and observes by the experimental validation that considering the splitting criteria for instances categorization can lead the classifier to perform better on withheld set.
Abstract: Fuzziness based divide and conquer (D&C) is a recently proposed strategy for promoting the classifiers (i.e., fuzzy classifiers) performance, where the amount of fuzziness quantity associated with each data point (i.e., both labeled and unlabeled) is considered as an important avenue to the empire for instance selection problem. This technique is regarded as a semi-supervised learning (SSL) technique, where different categories of instances are obtained by using fuzziness measure, and then the instances having less amount of fuzziness are incorporated into training set for improving the generalization ability of a classifier. This study proposes some effective methods and presents a novel algorithm for categorizing the instances into three groups that can effectively integrate with D&C strategy. It is observed by the experimental validation that considering the splitting criteria for instances categorization can lead the classifier to perform better on withheld set. Results on different classification data sets prove the effectiveness of proposed algorithm. 6
9 citations