The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
Knowledge about facts from the real world is very often uncertain, ambiguous and vague. As a result, the inference mechanism based on such facts generates uncertain and vague conclusions. In this paper we describe a formal model of knowledge representation for uncertain and vague domains. The model is based on the theory of high-level fuzzy Petri Nets (FPNs), and a fuzzy inheritance procedure that...
In this paper, a hybrid algorithm is proposed for designing feedforward neural networks. A genetic algorithm is used to tune the connections and parameters between the input layer and the hidden layer, and orthogonal transformation is applied to tune the connections and parameters between the hidden layer and the output layer. In this way, both the structure and parameters of a neural network can...
In this paper we consider a credit scoring problem. We compare three powerful credit scoring models: genetic programming (GP), backpropagation neural networks (BP) and support vector machines (SVM) when applied to this problem, then we give a combined model. The results show that the combined model produces good classification results.
Support vector machine (SVM) is a powerful tool to solve classification problems, this paper proposes a fast Sequential Minimal Optimization (SMO) algorithm for training one-class support vector regression (OCSVM), firstly gives a analytical solution to the size two quadratic programming (QP) problem, then proposes a new heuristic method to select the working set which leads to algorithm's faster...
The constructive neural networks based on the covering algorithms is suitable for large-scale data mining because it can be local processing and has little computational complexity, however, the local processing lows classification precision. In this paper, covering algorithms is firstly extended to kernel covering algorithms and we secondly construct a kind of finite mixture probabilistic model based...
In this paper, a new learning algorithm which encodes a priori information into feedforward neural networks is proposed for function approximation problem. The algorithm incorporates two kinds of constraints into single hidden layered feedforward neural networks, which are architectural constraints and connection weight constraints, respectively, from a priori information of function approximation...
A new algorithm called classification-rejection sphere support vector machines (C-R sphere SVMs) is proposed based on the human thoughts of recognition and support vector machine (SVM) technology for multi-class classification problems. The new algorithm constructs a classifying sphere for each class instead of a minimum sphere. Like human being, C-R sphere SVMs can not only classify the multi-class...
A novel recurrent generalized congruence neural network (RGCNN) is presented. Compared with traditional recurrent neural networks (RNNs), RGCNN has the following advantages: simple structure (4 layers), no time-consuming iterative derivative operations in updating weights, and fast convergence induced by modulo arithmetic of the generalized congruence neuron. Computer simulations on benchmark examples...
A time-delay recurrent neural network (TDRNN) model is proposed. TDRNN has a simple structure but far more "depth" and "resolution ratio" in memory by introducing the time-delay and recurrent mechanism. A dynamic recurrent back propagation algorithm is developed. The optimal adaptive learning rates are derived in the sense of discrete-type Lyapunov stability to guarantee the fast...
Independent component analysis (ICA), blind source separation (BSS) and related methods like blind source extraction (BSE) have been considered as a fundamental data analysis tool in the fields of neural network and signal processing. In this paper, we propose a robust algorithm based on a specific kurtosis value range that can extract a desired source signal as the first output signal with a specific...
Owning to the disadvantages of OLS in network structure optimization, authors put forward using Fisher ratio method to optimize the RBF centers, an orthogonal transform, and a forward selection search method are used to optimize structure. The simulation results show that the neural structure is simplified strongly, the converge precision and class separability are improved, and it is satisfied for...
This paper provides a novel multi-class classification algorithm, which combines adaptive resonance theory with support vector machine principle. It improves the one-against-one classification of support vector machine. The algorithm adopts adaptive resonance theory network to fuse the classifiers' results and does not adopt voting principle. When the outputs of classifiers approach zero and the algorithm...
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.