Showing papers in "Neurocomputing in 2003"

TL;DR: Experimental results with real data sets indicate that the combined model can be an effective way to improve forecasting accuracy achieved by either of the models used separately.

TL;DR: The experimental results show that SVM provides a promising alternative to stock market prediction and the feasibility of applying SVM in financial forecasting is examined by comparing it with back-propagation neural networks and case-based reasoning.

TL;DR: A popular SVM implementation is compared to 16 classification methods and 9 regression methods accessible through the software R by the means of standard performance measures and bias-variance decompositions which showed mostly good performances both on classification and regression tasks, but other methods proved to be very competitive.

TL;DR: The empirically study the usefulness of several simple performance measures that are inexpensive to compute (in the sense that they do not require expensive matrix operations involving the kernel matrix) for tuning SVM hyperparameters.

TL;DR: The experiment shows that SVM by feature extraction using PCA, KPCA or ICA can perform better than that without feature extraction, and among the three methods, there is the best performance in K PCA feature extraction; followed by ICA feature extraction.

TL;DR: Modifications of this algorithm that improve its learning speed are discussed and the new optimization methods are empirically compared to the existing Rprop variants, the conjugate gradient method, Quickprop, and the BFGS algorithm on a set of neural network benchmark problems.

TL;DR: The simulation shows that the SVMs experts achieve significant improvement in the generalization performance in comparison with the single SVMs models, and converge faster and use fewer support vectors.

TL;DR: The measure is efficient and faithful in characterizing spike timing reliability and produces smaller errors in the reliability estimate than the histogram-based measure based on the same number of trials.

TL;DR: For regression problems, based on scale space theory, it is demonstrated the existence of a certain range of σ, within which the generalization performance is stable, and an appropriate σ within the range can be achieved via dynamic evaluation.

TL;DR: The dynamic behaviour of the RNN is used to categorize input sequences into different specified classes and enables the user to assess efficiently the degree of reliability of the classification result.

TL;DR: In this paper, an extension to the case of SVMs with quadratic slack penalties is given and a simple approximation for the evidence is derived, which can be used as a criterion for model selection.

TL;DR: This paper proposes an improved version of the neocognitron and tests its ability using a large database of handwritten digits (ETL1) to improve the recognition rate and removes accessory circuits that were appended to the previous versions.

TL;DR: This article introduces the “Support Vector Classification-Regression” machine for K -class classification purposes ( K -SVCR), a new training algorithm with ternary outputs based on Vapnik's Support Vector theory, using a mixed classification and regression SV Machine (SVM) formulation.

TL;DR: An algorithm of evolving self-organizing map (ESOM), which features an evolving network structure and fast on-line learning, is presented, which is a promising computational model for on- line pattern analysis in real world problems.

TL;DR: The ANN and neuro-fuzzy approaches are used for handling the situations with scarce data, where the predictions are based on the upstream hydrological conditions only, and explicitly outperforming the linear statistical models for a longer prediction horizon.

TL;DR: The kernel trick is used to select from the data a relevant subset forming a basis in a feature space F, and it will turn out that the size of the basis is related to the complexity of the model.

TL;DR: A new three-term backpropagation algorithm is proposed in order to speed-up the weight adjusting process and generally out-performs the conventional algorithm in terms of convergence speed and the ability to escape from local minima.

TL;DR: The proposed method, which is extremely fast and terminates in 6 or 7 iterations, can handle classification problems in very high dimensional spaces, e.g. over 28,000, in a few seconds on a 400 MHz Pentium II machine.

TL;DR: A sliding-mode neural-network (SMNN) control system for the tracking control of an n rigid-link robot manipulator to achieve high-precision position control is presented, and no constrained conditions and prior knowledge of the controlled plant is required in the design process.

TL;DR: A newly developed method, particle swarm optimization (PSO) model, is adopted to train the perceptron and to predict the pollutant levels, establishing a new neural network model, PSO-based approach, established and completed.

TL;DR: In this article, the authors argue for a non-negative variant of the sparse coding model, based partly on neurophysiological grounds and partly on the intuitive understanding of parts-based representations.

TL;DR: It is shown that SVR can be regarded as a classification problem in the dual space and generalization bounds for classification can be extended to characterize the generalization performance of the proposed approach.

TL;DR: The investigations in this paper clearly indicate the crucial importance of SVM parameter optimisation as well as variable selection in order to develop statistical models with good predictive capabilities on external test sets when using SVM regression.

TL;DR: A new strategy for adaptively and autonomously constructing a multi-hidden-layer feedforward neural network (FNN) that adds both new hidden units and new hidden layers one at a time when it is determined to be needed is introduced.

TL;DR: This paper focuses on the blind separation of n sources from m mixtures when the underlying system is underdetermined, as applied to mixtures with only attenuations and delays, using synthetic mixtures simulating real acoustic scenarios without reverberation.

TL;DR: A neural approach is used to model the rainfall-runoff process when different time step durations have to be considered in reservoir management.

TL;DR: A novel method for pattern recognition using discrete Fourier transforms on the global pulse signal of a pulse-coupled neural network (PCNN) to achieve scale- and translation-independent recognition for isolated objects.

TL;DR: Evidence was found for millisecond delays that could only be accounted for by assuming temporal structure in the constituent neurons’ spike trains, and the method was also applied to higher-order patterns.

TL;DR: In this paper, the authors combine multiple KIII sets into the KIV model, which approximates the operation of the basic vertebrate forebrain together with the basal ganglia and motor systems.

TL;DR: A modified estimator is developed for linear characterization of neurons when spikes arise from a leaky integrate-and-fire mechanism, and it is shown that spiking dynamics may account for changes observed in the receptive fields measured at different contrasts.