Moisés Salmerón

Improved RAN sequential prediction using orthogonal techniques

Abstract A new learning strategy for time-series prediction using radial basis function (RBF) networks is introduced. Its potential is examined in the particular case of the resource allocating network model, although the same ideas could apply to extend any other procedure. In the early stages of learning, addition of successive new groups of RBFs provides an increased rate of convergence. At the same time, the optimum lag structure is determined using orthogonal techniques such as QR factorization and singular value decomposition (SVD). We claim that the same techniques can be applied to the pruning problem, and thus they are a useful tool for compaction of information. Our comparison with the original RAN algorithm shows a comparable error measure but much smaller-sized networks. The extra effort required by QR and SVD is balanced by the simplicity of only using least mean squares for the iterative parameter adaptation.

Web newspaper layout optimization using simulated annealing

The Web newspaper pagination problem consists of optimizing the layout of a set of articles extracted from several Web newspapers and sending it to the user as the result of a previous query. This layout should be organized in columns, as in real newspapers, and should be adapted to the client Web browser configuration in real time. This paper presents an approach to the problem based on simulated annealing (SA) that solves the problem on-line, adapts itself to the clients computer configuration, and supports articles with different widths.

A new radial basis function networks structure: application to time series prediction

Describes a structure to create a RBF neural network. This structure has 4 main characteristics. The first one is that the special RBF network architecture uses regression weights to replace the constant weights normally used. These regression weights are assumed to be functions of input variables. The second characteristic is the normalization of the activation of the hidden neurons (weighted average) before aggregating the activations, which, as observed by various authors, produces better results than the classical weighted sum architecture. The third aspect is that a new type of nonlinear function is proposed, the pseudo-gaussian function (PGBF). With this, the neural system gains flexibility, as the neurons possess an activation field that does not necessarily have to be symmetric with respect to the centre or to the location of the neuron in the input space. In addition to this new structure, we propose, as the fourth and final feature, a sequential learning algorithm, which is able to adapt the structure of the network, with this, it is possible to create new hidden units and also to detect and remove inactive units.

Time series prediction using ICA algorithms

We propose a new method for volatile time series forecasting using independent component analysis (ICA) algorithms and Savitzky-Golay filtering as preprocessing tools. The preprocessed data will be introduce in a based radial basis functions (RBF) artificial neural network (ANN) and the prediction result will be compared with the one we get without these preprocessing tools or the classical principal component analysis (PCA) tool

New Model for Time-Series Forecasting Using RBFs and Exogenous Data

In this paper we present a new model for time-series forecasting using Radial Basis Functions (RBFs) as a unit of ANN ’s (Artificial Neural Networks), which allows the inclusion of exogenous information (EI) without additional preprocessing. We begin summarizing the most well known EI techniques used ad hoc, i.e. PCA or ICA; we analyse advantages and disadvantages of these techniques in time-series forecasting using Spanish banks and companies stocks. Then we describe a new hybrid model for time-series forecasting which combines ANN ’s with GA (Genetic Algorithms); we also describe the possibilities when implementing on parallel processing systems.

ANNs and GAs for predictive control of water supply networks

This paper describes a procedure for controlling the water supply system. The controller uses a neural network to predict the water demand levels and a genetic algorithm to determine the feasible operation points in an optimal strategy that is based on dynamic programming. The controller has been executed in parallel in a cluster of computers. This has allowed not only the determination of the control commands in the required times but also the improvement of the control procedure performances.

Parallel combinatorial optimization with evolutionary cooperation between processors

An evolutionary computation approach is used to learn online the rules that allow the processors in a parallel platform to cooperate by interchanging the local optima that they find while they concurrently explore different zones of the solution space. The cooperation of processors can greatly benefit the resolution of combinatorial optimization problems by decreasing their runtimes, by increasing the quality of the solutions obtained, or both. Moreover, as parallel computers are more and more accessible, the application of parallel processing to solve these problems becomes a practical and interesting alternative. As an example, a parallel optimization algorithm based on Boltzmann Machine has been used for a detailed description and evaluation of the proposed cooperation approach.

Hybridizing Genetic Algorithms with ICA in Higher Dimension

In this paper we present a novel method for blindly separating unobservable independent component signals from their linear mixtures, using genetic algorithms (GA) to minimize the nonconvex and nonlinear cost functions. This approach is very useful in many fields such as forecasting indexes in financial stock markets where the search for independent components is the major task to include exogenous information into the learning machine. The GA presented in this work is able to extract independent components with faster rate than the previous independent component analysis algorithms based on Higher Order Statistics (HOS) as input space dimension increases showing significant accuracy and robustness.

On-Line Optimization of Radial Basis Function Networks with Orthogonal Techniques

In this paper the QR-cp factorization and Singular Value Decomposition (SVD) matrix numerical procedures are used for the optimization of the structure of Radial Basis Function (RBF) neural networks—that is, the best number of input nodes and also the number of neurons within the network. We study the application domain of time series prediction and demonstrate the superior performance of our method for on-line prediction of a well known chaotic time series. A new strategy that consists of the initial allocation of successive groups of nodes is also suggested, since it leads to initially faster learning.

Expert Mutation Operators for the Evolution of Radial Basis Function Neural Networks

This paper compares some mutationoperators containing expert knowledge about the problem of optimizing the parameters of a Radial Basis Function Neural Network. It is shown that the expert knowledge is not always able to improve the results obtained by a blind evolutionary algorithm, and that the final results depend strongly on how the expert knowledge is utilized.