Federated learning trains a shared model across many clients without moving raw data, while continual learning learns a stream of tasks and mitigates catastrophic forgetting. Continual federated learning combines these goals but is challenged by non-IID label skew and forgetting under evolving client data. We propose GAN-CFL (Generative Adversarial Networks-based Continual Federated Learning) to tackle these challenges. GAN-CFL enables clients to learn from new data, without storing historical data, and effectively adapts to non-IID data distributions. GAN-CFL enhances data heterogeneity by generating synthetic data to augment real datasets and mitigates catastrophic forgetting across multiple clients by incorporating elastic weight consolidation algorithm. In this framework, the generator produces synthetic images, while the discriminator classifies both real and generated images. The trained discriminator is then used as a classifier on real data to provide accuracy metrics. The global model aggregates local weights from clients to optimize overall performance. We evaluate GAN-CFL on six datasets, MNIST, K-MNIST, Fashion-MNIST, EMNIST-letters, EMNIST-Balanced, and CIFAR-10 for experiments involving up to 100 clients. Our model is compared to a centralized learning method for ablation analysis. The results show that GAN-CFL outperforms existing methods in classification accuracy.

The graph convolutional network (GCN) has demonstrated effectiveness well in the text classification task. However, inadequate handling of uncertainty in prediction results exists due to the under-utilization of text features extracted by a single deep-learning model. To mitigate the potential risk of text misclassification, we proposed an enhanced GCN model for text classification based on three-way decision, incorporating shadowed set theory (3WD-GCN). In this approach, we first employ GCN as a primary classifier to handle textual data, obtaining the initial predicted results and the membership matrix. Depending on the idea of processing in threes, these results were divided into acceptance, rejection, and subdivision regions, respectively. For the subdivision region, we introduce SVM as a secondary classifier to process objects with poor conformability and distinguishability, which can reduce the uncertainty of prediction results and improve the overall performance of text classification. A series of experiments based on several benchmark datasets extensively evaluated the proposed method. The results demonstrate the validity of the approach and show a significant improvement over popular baseline text classification models.

Addressing uncertainty issues is a significant challenge in decision-making. Soft set theory is designed to assist in complex decision-making scenarios where multiple approximations are involved. Those approximations, represented as parametrized sets in soft sets, could provide decision-makers with more informed choices when integrated effectively. However, con-flicts among distinct approximations make the integration chal-lenging. To address this issue, we propose a game-theoretic soft set model based on a set-oriented perception. This model effectively manages the merging of approximation sets by dividing the universe into overlapping and conflicting regions and employing tailored strategies for each. Experimental results indicate that the model not only can achieve a balance among various parameters or conflicting decision goals, but also improves approximation performance across accuracy, precision, recall, and Fl-score.

Machine learning has been applied to classify chest X-ray images into pneumonia-positive and pneumonia-negative classes to allow an early diagnosis and support medical experts’ decision about pneumonia. However, the previous attempts in this literature focus on binary classification that may not consider the possibility of uncertain information in chest X-ray images, forcing the system to make a definite decision on every instance. This may lead to the inaccurate classification of doubtful X-ray images. This research approaches Game-theoretic rough sets (GTRS) to determine three-way decisions, such as X-ray images are classified into three disjoint classes that are developed using a threshold pair. The first two classes lead to a certain decision (i.e., either pneumonia-positive or pneumonia-negative). The remaining class covers X-ray images that lack the crucial information to make inferences about them. GTRS obtains the suitable threshold pair by formulating a trade-off between accuracy and coverage criteria of the proposed model. We achieve a 96.25% accuracy score while covering 64.01% of the test set for certain decision. The experiment results show that the proposed model secures a better classification performance than 0.5-probabilistic rough sets, Pawlak’s rough sets, and machine learning models.

Image processing techniques are essential in understanding and analyzing information based on pictorial data. The three-way decision is a natural extension of the widely accepted binary-decision model with an added third option. We propose a three-way image classification model in this article. The model integrates techniques of image processing and three-way decision for classification problems. The proposed model is performed in six steps, including HSV and extracted textual local binary pattern (LBP) information, which are: principal component analysis (PCA) feature extraction; decision table creation; classification space development; positive, boundary, and negative sets initialization and optimization; boundary reduction; and test verification. The proposed three-way decision model has been tested with a corn grain dataset under challenges such as (e.g. color component variation, surface pattern variation or so on), that achieved 99 percent performance accuracy.

Shonosuke Ishiwatari, Jingtao Yao, Shujie Liu, Mu Li, Ming Zhou, Naoki Yoshinaga, Masaru Kitsuregawa, Weijia Jia. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2017.

In this paper, we are concerned with a problem related to granular computing with game theoretic intuitionistic fuzzy rough sets(GTIFRS). The game theoretic intuitionistic fuzzy rough sets is defined over compatibility based intuitionistic fuzzy relation. The imprecise information obtained, is considered to be composed of intuitionistic fuzzy granules. The basis of granulation is splitting intervals into several sub-intervals. A competitive situation arises due to movement of granules which is controlled by threshold parameters.

In this paper, we propose a mechanism for cancer cell differentiation therapy based on a game-theoretic rough set model with intuitionistic fuzzy set. We introduce an intuitionistic fuzzy relation with the help of fuzzy compatibility relation. The approximation space classification is controlled with a pair of intuitionistic fuzzy thresholds. It provides a flexibility to set up tolerance level depending on different requirements. The threshold parameters act as players and pay-offs are determined based on the shifting of equivalent classes from boundary region to positive or negative region.

The realization of the Web as a common platform, medium, and interface for supporting human activities has attracted many researchers to the study of Web-based support systems (WSS). An important branch of WSS is Web-based decision support systems that provide intelligent support for making effective decisions in different domains. We focus on decision making in Web-based medical decision support systems (WMDSS). Uncertainty is a critical factor that affects decision making and reasoning in the medical field. A three-way decision-making approach is an effective and better choice to lessen the effects of uncertainty. It provides the provision for delaying certain or definite decisions in situations that lack sufficient evidence or accurate information in reaching certain conclusions. Particularly, the option of deferment decisions is added in this approach that provides the flexibility to further examine and investigate the uncertain and doubtful cases. The game-theoretic rough set (GTRS) model is a recent development in rough sets that can be used to determine the three rough set regions in the probabilistic rough sets framework by determining a pair of thresholds. The three regions are used to obtain three-way decision rules in the form of acceptance, rejection, and deferment rules. In this paper, we extend the GTRS model to analyze uncertainty involved in medical decision making. Experimental results with a GTRS-based approach on different health care datasets suggest that the approach may improve the overall quality of decision making in the medical field, as well as other fields. It is hoped that the incorporation of a GTRS component in WMDSS will enrich and enhance its decision-making capabilities.

Web-based support systems (WSS) assist human activities with the modern Web technology. An important branch of WSS is Web-based decision support systems that provide intelligent support for decision making tasks. We focus on decision making in Web-based medical decision support systems (WMDSS) that can provide support for making diagnosis and treatment decisions. The use of game-theoretic rough set (GTRS) component in WMDSS is explored and investigated for this purpose. The GTRS is a recent development in rough sets that takes advantages from data analysis capabilities of rough sets complimented with decision analysis abilities of game theory. The GTRS may be used to obtain rough sets based three-way or ternary decisions by determining a pair of threshold values. Demonstrative example suggests that the GTRS may be considered as an alternative decision making model and component in building WMDSS for providing decision support.

This special issue of Fundamenta Informaticae contains a selection of papers initially presented at the 6th International Conference on Rough Sets and Knowledge Technology (RSKT'11) held during October 8-11, 2011 in Banff, Canada.RSKT is an international scientific conference series that has been held every year since 2006.The conferences serve as a major forum that brings researchers and industry practitioners together to discuss and deliberate on fundamental issues of knowledge processing and management and knowledge-intensive practical solutions in the current knowledge age.Experts from around the world meet to present state-of-the-art scientific results, to nurture academic and industrial interaction, and to promote collaborative research in rough sets and knowledge technology.We initially had twelve papers invited.After rigorous review, eight papers were selected to be included in this issue.They are substantially extended versions of respective conference papers.Each paper was review by three domain experts and went through at least two rounds of revisions.

Granular computing, as a new and rapidly growing paradigm of information processing, has attracted many researchers and practitioners. Granular computing is an umbrella term to cover any theories, methodologies, techniques, and tools that make use of information granules in complex problem solving. The aim of this paper is to review foundations and schools of research and to elaborate on current developments in granular computing research. We first review some basic notions of granular computing. Classification and descriptions of various schools of research in granular computing are given. We also present and identify some research directions in granular computing.

The decision-theoretic rough set (DTRS) model considers costs associated with actions of classifying an equivalence class into a particular region. With DTRS, one may make informative decisions in the form of three-way decisions. Current research mai

This article investigates the Game-theoretic Rough Set (GTRS) model and its capability of analyzing a major decision problem evident in existing probabilistic rough set models. A major challenge in the application of probabilistic rough set models is

In the ideal case of complete knowledge, for each property Pi(such as “high fever”, “headache”, etc.), we know the exact set Siof all the objects that satisfy this property. In practice, we usually only have partial knowledge. In this case, we only know the set Siof all the objects about which we know that Piholds and the set Siabout which we know that Pimay hold (i.e., equivalently, that we have not yet excluded the possibility of Pi). This pair of sets is called a set interval. Based on the knowledge of the original properties, we would like to describe the set S of all the values that satisfy some combination of the original properties: e.g., high fever and headache and not rash. In the ideal case when we know the exact set Siof all the objects satisfying each property, it is sufficient to apply the corresponding set operation (composition of union, intersection, and complement) to the known sets Si. In this paper, we describe how to compute the class S of all possible sets S.

Many information retrieval and machine learning methods have not evolved in order to be applied to the Web. Two main problems in applying some machine learning techniques for Web mining are the dynamic and ever-changing nature of Web data and the sheer size of possible dimensions that this data could portray. One such technique, self-organizing maps (SOMs), have been enhanced to deal with these two problems individually. The growing hierarchical self-organizing map can adapt to the dynamic data present on the Web by changing its topology according to the amount of change in input size. In addition, it reduces local dimensionality by splitting features into levels. We extend this model by including bidirectional update propagation over the levels of the hierarchy. We demonstrate the effectiveness of the new approach with a Web-based news coverage example.

A partition-based framework is presented for a formal study of classification problems. An information table is used as a knowledge representation, in which all basic notions are precisely defined by using a language known as the decision logic language. Solutions to, and solution space of, classification problems are formulated in terms of partitions. Algorithms for finding solutions are modelled as searching in a space of partitions under the refinement order relation. We focus on a particular type of solutions called conjunctively definable partitions. Two level-wise methods for decision tree construction are investigated, which are related to two different strategies: local optimization and global optimization. They are not in competition with, but are complementary to each other. Experimental results are reported to evaluate the two methods.

WIC Canada promotes the collaboration between Canadian researchers on Web Intelligence, facilitates exchange with other WIC centers. WIC Canada researchers work on a diversity of WI related research areas: foundations of Web Intelligence, Web-based support systems, Bayesian networks, and IntelligentWeb Information Systems (IWIS). Their results have appeared in reputable journals and the annual IEEE/WIC/ACM International Conference onWeb Intelligence. In the next few years, WIC Canada will focus on promotingWeb Intelligence research, attracting new members, and forming new centres. The WIC Canada will play the role of coordinating those activities. WIC Canada needs your input for its growth. Your participation will be greatly appreciated. All questions and suggestions should be directed to our Co-ordinator, Dr. Jingtao Yao, at [email protected].

Finding all distinct matchings of the query tree pattern is the core operation of XML query evaluation. The existing methods for tree pattern matching are decomposition-matching-merging processes, which may produce large useless intermediate result or require repeated matching of some sub-patterns. We propose a fast tree pattern matching algorithm called TreeMatch to directly £nd all distinct matchings of a query tree pattern. The only requirement for the data source is that the matching elements of the non-leaf pattern nodes do not contain sub-elements with the same tag. The TreeMatch does not produce any intermediate results and the £nal results are compactly encoded in stacks, from which the explicit representation can be produced ef£ciently.

The concept of Web-based information retrieval support systems (WIRSS) is introduced. The needs for WIRSS are shown by a detailed case study of existing research article indexing and citation analysis systems, such as current content, DBLP, science citation index and CiteSeer. The objective of WIRSS is to build new and effective research tools for scientists to access, explore and use information on the Web, which may lead to improved research productivity and quality.

Goodness-of-fit is the most popular criterion for neural network time series forecasting. In the context of financial time series forecasting, we are not only concerned at how good the forecasts fit their targets, but we are more interested in profits. In order to increase the forecastability in terms of profit earning, we propose a profit based adjusted weight factor for backpropagation network training. Instead of using the traditional least squares error, we add a factor which contains the profit, direction, and time information to the error function. The results show that this new approach does improve the forecastability of neural network models, for the financial application domain.