This thesis investigates feedforward neural networks in the context of classification tasks with respect to the detection of patterns that do not belong to the same categories of patterns used to train the network. This refers to the problem of the detection and/or rejection of spurious or novel patterns. In particular, the multilayer perceptron network (MLP) trained with the backpropagation algorithm is examined in this respect and different strategies for improving its performance in the detection of spurious patterns are considered. The problem is investigated from different points of view that vary from the modification of the multilayer perceptron network with different configurations that make it more intrinsically able to detect spurious information, to the introduction of novel auxiliary mechanisms which, when integrated with the MLP network, can provide an overall enhancement in the system's rejection capabilities. These different network configurations are examined with respe...

Accurate prediction of ionospheric parameters is crucial for telecommunication companies. These parameters strongly rely on solar activity. In this paper, we analyze the use of neural networks for sunspots time series prediction. Three types of models are tested and experimental results are reported for a particular sunspots time series: the IR5 index.

Abstract not found

Abstract not found

The capability of recurrent neural networks of approximating functions from lists of real vectors to a real vector space is examined: Any measurable function can be approximated in probability. Additionally, bounds on the resources sufficient for an approximation can be derived in interesting cases.

¯ The objective of this paper is to present experiments and discussions of how some neural network algorithms can help the phoneme recognition with mixture density hidden Markov models (MDHMMs). In MDHMMs the modeling of the stochastic observation processes associated with the states is based on the estimation of the probability density function of the short-time observations in each state as a mixture of Gaussian densities. The Learning Vector Quantization (LVQ) is used to increase the discrimination between dioeerent phoneme models both during the initialization of the Gaussian codebooks and during the actual MDHMM training. The Self-Organizing Map (SOM) is applied to provide a suitably smoothed mapping of the training vectors to accelerate the convergence of the actual training. The obtained codebook topology can also be exploited in the recognition phase to speed up the calculations to approximate the observation probabilities. The experiments with LVQ and SOMs show reductions both...

. Two types of artificial neural networks are introduced for the robust classification of spatio-temporal sequences. The first network is the Adaptive Spatio-Temporal Recognizer (ASTER), which adaptively estimates the confidence that a (variable length) signal of a known class is present by continuously monitoring a sequence of feature vectors. If the confidence for any class exceeds a threshold value at some moment, the signal is considered to be detected and classified. The nonlinear behavior of ASTER provides more robust performance than the related dynamic time warping algorithm. ASTER is compared with a more common approach wherein a self-organizing feature map is first used to map a sequence of extracted feature vectors onto a lower dimensional trajectory, which is then identified using a variant of the feedforward time delay neural network. The performance of these two networks is compared using artificial sonograms as well as feature vectors strings obtained from short-duration...

Contents Page List of Figures ..................................................................... ..... v List of Tables ..................................................................... ......vii Acknowledgments...................................................... ..............viii 1.- Introduction ..................................................................... ... 1 Objective of the Thesis Research............................................ 3 2.- Foundations ..................................................................... ... 5 a) The Dynamic Nature of Speech .......................................... 6 b) The Use of Spectrograms for Speech Recognition ....................11 c) Speech Recognition Schools of Thought................................14 c.1) The Template-Based Approach .................................15 c.2) The Feature-Based Approach .......................

The problem of the rejection of patterns not belonging to identified training classes is investigated with respect to multilayer perceptron networks (MLPs). The reason for the inherent unreliability of the standard MLP in this respect is explained and some mechanisms for the enhancement of its rejection performance are considered. Two network configurations are presented as candidates for a more reliable structure and are compared to the so-called "negative training" approach. The first configuration is an MLP which uses a Gaussian as its activation function and the second is an MLP with direct connections from the input to the output layer of the network. The networks are examined and evaluated both through the technique of network inversion and through practical experiments in a pattern classification application. Finally, the model of radial basis function networks (RBFs) is also considered in this respect and its performance is compared to that attained with the other networks desc...

. There have been a number of recent papers on information theory and neural networks, especially in a perceptual system such as vision. Some of these approaches are examined, and their implications for neural network learning algorithms are considered. Existing supervised learning algorithms such as Back Propagation to minimize mean squared error can be viewed as attempting to minimize an upper bound on information loss. By making an assumption of noise either at the input or the output to the system, unsupervised learning algorithms such as those based on Hebbian (principal component analysing) or anti-Hebbian (decorrelating) approaches can also be viewed in a similar light. The optimization of information by the use of interneurons to decorrelate output units suggests a role for inhibitory interneurons and cortical loops in biological sensory systems. 1. Introduction Almost as soon as Shannon first formulated his `Mathematical Theory of Communication' [1], psychologists and physiol...