2,979 publications from this institution
The concept of temporally sensitive fuzzy neural networks is introduced based on combining the basic ideas of logic-based neurocomputing with the concept of temporally sensitive connections of neural networks. This new class of neural networks helps address two main issues arising in time-dependent modeling environments. Firstly, these neural networks capture the underlying logical fabric of the problem and, secondly, they provide a useful insight into the temporal nature of the modeling environment. The authors show that the continuously changeable temporal environment gives rise to a logical transformation of the introduced model. This transformation is implemented by triggering from its original AND-like nature to an OR-like version, with this triggering regarded as a function of time. This paper discusses fuzzy decision-making in detail, particularly real estate problem solving.
Software Engineering is inherently knowledge intensive. Software processes and products are human centered. The technology of Computational Intelligence (CI) intensively exploits various mechanisms of interaction with humans and processes domain knowledge with intent of building intelligent systems. As commonly perceived, CI dwells on three highly synergistic technologies of neural networks, fuzzy sets (or granular computing, in general) and evolutionary optimization. As the software complexity grows and the diversity of software systems skyrocket, it becomes apparent that there is a genuine need for a solid, efficient, designer-oriented vehicle to support software analysis, design, and implementation at various levels. The research agenda makes CI a highly compatible and appealing vehicle to address the needs of knowledge rich environment of Software Engineering. The objective of this study is to identify and discuss synergistic links emerging between Software Engineering and Computational Intelligence. We show how CI --- based models contribute to the methodology of constructing models of software processes and products. Several selected examples (including software cost estimation, quality, and software measures) are included.
In this paper, an embedded feature selection based on variational relevance vector machines is proposed to simultaneously perform feature selection and model construction. With the settings of specific hierarchical priors over the parameters of an automatic relevance determination kernel (ARDK) function, an approximate posterior distribution over these parameters is here derived and expressed as a multivariate Gaussian distribution, in which a first-order Taylor expansion-based Laplace approximation with respect to the parameters is introduced into the variational inference procedure. The posterior distributions, rather than generic pointwise estimates, over the rest of parameters of the model are also derived. The proposed method can simultaneously select relevant features and samples by adjusting the parameters of ARDK and the weighting vector, respectively. To verify the effectiveness of the proposed method, a synthetic dataset and a number of benchmark datasets, as well as a practical industrial dataset, are employed to solve the regression and classification problems. These experimental results indicate that the proposed method supports the mechanisms of feature selection and model construction while maintaining prediction performance, particularly in an industrial environment.
Attacks over the Internet are becoming more and more complex and sophisticated. How to detect security threats and measure the security of the Internet arises a significant research topic. For detecting the Internet attacks and measuring its security, collecting different categories of data and employing methods of data analytics are essential. However, the literature still lacks a thorough review on security-related data collection and analytics on the Internet. Therefore, it becomes a necessity to review the current state of the art in order to gain a deep insight on what categories of data should be collected and which methods should be used to detect the Internet attacks and to measure its security. In this paper, we survey existing studies about security-related data collection and analytics for the purpose of measuring the Internet security. We first divide the data related to network security measurement into four categories: 1) packet-level data; 2) flow-level data; 3) connection-level data; and 4) host-level data. For each category of data, we provide a specific classification and discuss its advantages and disadvantages with regard to the Internet security threat detection. We also propose several additional requirements for security-related data analytics in order to make the analytics flexible and scalable. Based on the usage of data categories and the types of data analytic methods, we review current detection methods for distributed denial of service flooding and worm attacks by applying the proposed requirements to evaluate their performance. Finally, based on the completed review, a list of open issues is outlined and future research directions are identified.
In this paper, we introduce the design methodology of interval type-2 fuzzy neural networks (IT2FNN). And to optimize the network we use a real-coded genetic algorithm. IT2FNN is the network of combination between the fuzzy neural network (FNN) and interval type-2 fuzzy set with uncertainty. The antecedent part of the network is composed of the fuzzy division of input space and the consequence part of the network is represented by polynomial functions. The parameters such as the apexes of membership function, uncertainty parameter, the learning rate and the momentum coefficient are optimized using genetic algorithm (GA). The proposed network is evaluated with the performance between the approximation and the generalization abilities.
This paper follows a companion paper (Stochastica 8 (1984), 99-145) in which we gave the state of the art of the theory of fuzzy relation equations under a special class of triangular norms. Here we continue this theory establishing new results under lower and upper semicontinuous triangular norms and surveying on the main theoretical results appeared in foregoing papers. Max-t fuzzy equations with Boolean solutions are recalled and studied. Many examples clarify the results established.
In the plethora of conceptual and algorithmic developments supporting system modeling, we encounter growing challenges associated with the complexity of systems, diversity of available data and a variety of requests imposed on the quality of the models. The accuracy of models is important. At the same time, the interpretability and explainability of models are equally important and of high practical relevance. We advocate that the level of abstraction at which models are constructed (and which could be flexibly adjusted), is conveniently realized through Granular Computing. Granular Computing is concerned with the development and processing information granules - formal entities that facilitate a way of organizing and representing knowledge about the available data and relationships existing there. This study identifies the principles of Granular Computing, shows how information granules are constructed and subsequently used in the realization of models.
In this paper, we develop a comprehensive identification scheme for fuzzy relation-based neural networks (FRNNs). The proposed hybrid development approach combines the optimization technology of genetic algorithms (GAs) and an improved complex method introduced in the previous studies on fuzzy modeling. The structure of the FRNNs revolves around a collection of fuzzy rules and involves two types of fuzzy inference schemes. The taxonomy of these schemes relates to the format of the conclusion part of these rules (being either constants or linear functions). The optimization of the network deals with a number of essential parameters as well as the underlying learning mechanisms (e.g., apexes of membership functions, learning rates, and momentum coefficients). The hybrid identification approach helps achieve global optimization (when using GAs) and assure local convergence (that results from the use of the improved complex method). During the identification process, we are guided by a weighted objective function (performance index) in which a weighting factor is introduced to achieve a sound balance between approximation and generalization capabilities of the resulting model. The proposed identification method is applied to nonlinear processes (data) such as gas furnace process data and emission process data form a gas turbine power plant. The obtained experimental results show that the proposed networks exhibit high accuracy and generalization capabilities.
