Bonifacio Martín-del-Brío

Self-organizing neural networks for the analysis and representation of data: some financial cases

Many recent papers have dealt with the application of feedforward neural networks in financial data processing. This powerful neural model can implement very complex nonlinear mappings, but when outputs are not available or clustering of patterns is required, the use of unsupervised models such as self-organizing maps is more suitable. The present work shows the capabilities of self-organizing feature maps for the analysis and representation of financial data and for aid in financial decision-making. For this purpose, we analyse the Spanish banking crisis of 1977–1985 and the Spanish economic situation in 1990 and 1991, making use of this unsupervised model. Emphasis is placed on the analysis of the synaptic weights, fundamental for delimiting regions on the map, such as bankrupt or solvent regions, where similar companies are clustered. The time evolution of the companies and other important conclusions can be drawn from the resulting maps.

Sensor linearization with neural networks

A procedure for extending the linear range of an arbitrary sensor is proposed. The process is carried out by a neural network which compensates the sensor nonlinear characteristic. A negative temperature coefficient resistor sensor is used as an application example of the procedure, and its implementation in low-resolution microcontrollers is analyzed.

Use of an AC self-excited switched reluctance generator as a battery charger

We analyzed an AC self-excited switched reluctance generator used as a battery charger, using a second- order nonlinear oscillator model. Generator capacity can be maintained under variable rotor speed by adapting the external capacitance. The AC voltage generated is rectified and adapted with a DC-DC converter to battery voltage. To keep the flux machine under control, the rectified bus voltage should be changed according to oscillating frequency, following a constant voltage/frequency ratio. The simplicity of the system makes it an adequate wind generator for battery charging in isolated locations.

Computational intelligence tools for next generation quality of service management

In this paper we explore the interest of computational intelligence tools in the management of heterogeneous communication networks, specifically to predict congestion, failures and other anomalies in the network that may eventually lead to degradation of the quality of offered services. We show two different applications based on neural and neuro-fuzzy systems for quality of service (QoS) management in next generation networks for voice and video service over heterogeneous Internet protocol (V2oIP) services. The two examples explained in this paper attempt to predict the communication network resources for new incoming calls, and visualizing the QoS of a communication network by means of self-organizing maps.

Feature Map Architectures for Pattern Recognition: Techniques for Automatic Region Selection

In the present paper we address several issues about data processing with Kohonen maps. First, this neural system and multidimensional scaling are compared. Next, several unsupervised techniques for delimiting domains on a Kohonen map are described. The different procedures are illustrated by using real-world data.

Quality of Life Evaluation of Elderly and Disabled People by Using Self-Organizing Maps

Elderly people usually have some disabilities that get worst with the years. Many times these disabilities difficult the tasks carried out in a normal independent life, as is the case with home tasks. In addition, about one fourth of the household accidents happen in the kitchen. Within the framework of a European project ---Easyline plus-, we have developed a tool to evaluate the quality of life of elderly people based on kitchen activity, extracted from data provided by the appliances. Such a tool has found to be very useful for social carers in order to monitor elderly activity, and as an objective support for diagnosis of the evolution of the personal abilities and autonomy of the user.

3D FEM characterization of a switched reluctance motor from direct experimental determination of the material magnetization curve

Analytical characterization of a switched reluctance motor (SRM) by means of magnetic circuit theory can be a tedious task. Specific software based on the finite elements method (FEM) for solving non linear magneto-static problems has proved very useful for this characterization. However, to solve the Maxwell equations it is necessary first to know the magnetization curve of the material used for building the motor, which may not be available in the program materials library. A simple and low cost method for a direct determination of the first magnetization curve of a magnetic material is presented in this communication. Through one single test of short duration, (which allows to reach high values of the magnetic field strength while keeping the material temperature practically constant), we obtain a B-H curve that fits correctly to the samples obtained by a traditional method requiring the identification of multiple hysteresis loops. This B-H curve has been included in a FEM program to determine the flux linkage-current-position characteristic of a SRM prototype. The results again fit those obtained experimentally, confirming the validity of the proposed method.

Neural networks for QoS network management

In this paper we explore the interest of computational intelligence tools in the management of heterogeneous communication networks, specifically to predict congestion, failures and other anomalies in the network that may eventually lead to degradation of the quality of offered services. We show two different applications based on neural and neurofuzzy systems for Quality of Service (QoS) management in next generation networks for V2oIP services. The two examples explained in this paper attempt to predict the communication network resources for new incoming calls, and visualizing by means of self-organizing maps the QoS of a communication network.

Microcontroller implementation of a voice command recognition system for human-machine interface in embedded systems

Recent developments in signal processing, pattern recognition, neural networks and powerful (low-cost) embedded processors allow including speech processing for human-machine interaction in consumer equipment. In this paper we present a compact system for vowels and small vocabulary recognition, its orientation being consumer applications, where cost is of paramount importance. In this case, the speech recognition method must be as simple as possible, in such a way that could be programmed onto a standard microcontroller device, with the typical requirements of low range embedded application, where memory (kilobytes of RAM) and computer power (some MHz) are very limited. In this work we show both, algorithms and their implementation onto a low-cost microcontroller, and we study the final system performance

Smart sensing with adaptive analog circuits

This work shows the design and application of a mixed-mode analog-digital neural network circuit for sensor conditioning applications. The proposed architecture provides a high extension of the linear range for non-linear output sensors as negative temperature coefficient resistors (NTC) or giant magnetoresistive (GMR) angular position sensors, by using analog current-mode circuits with digital 8-bit weight storage. We present an analog current-based neuron model with digital weights, showing its architecture and features. By modifying the algorithm used in off-chip weight fitting, main differences of the electronic architecture, compared to the ideal model, are compensated. A small neural network based on the proposed architecture is applied to improve the output of NTC thermistors and GMR sensors, showing good results. Circuit complexity and performance make these systems suitable to be implemented as on-chip compensation modules.