Abdennasser Chebira

Data driven multiple neural network models generator based on a tree-like scheduler

In this paper we describe a new penalty-based model selection criterion for nonlinear models which is based on the influence of the noise in the fitting. According to Occams razor we should seek simpler models over complex ones and optimize the trade-off between model complexity and the accuracy of a models description to the training data. An empirical derivation is developed and computer simulations for multilayer perceptron with weight decay regularization are made in order to show the efficiency and robustness of the method in comparison with other well-known criteria for nonlinear systems.

Non Linear Process Identification Using a Neural Network Based Multiple Models Generator

Identification of non-linear systems is an important task for model based control, system design, simulation, prediction and fault diagnosis. In real world applications, strong linearity and large number of related parameters make the realization of those steps challenging, and so, the identification task difficult. Recently, a number of works based on Multiple Modelling have been proposed to avoid difficulties related to non-linearity. In this paper we use an Artificial Neural Network based data driven Multiple Model generator, that we called T-DTS (Treelike Divide To Simplify), for non-linear systems identification. T-DTS reduces modeling complexity on both data and processing levels. The efficiency of such approach has been analyzed trough two applications dealing with none-linear process identification. Experimental results validating our approach have been reported.

Hybrid neural-based decision level fusion architecture: application to road traffic collision avoidance

A hybrid decision level architecture for a road collision risks avoidance system is presented. The goal of the decision level is to clas- sify the behavior of the vehicles observed by a smart system or vehicle. The knowledge of vehicle behavior enables the best management of the smart system resources. The association of a model to each observed vehicle mainly enables the limitation of inference and of the set of actions to be activated; thus the interactions between system levels can be more intelligent. The decision level of this architecture is composed of a neural classifier, which is associated to a numerical classifier. Each of these classifiers provides decisions that are expressed within the framework of fuzzy theory. An optimal fusion policy is reached using the functional neural network tool.

Multi-neural Network Approach for Classification of Brainstem Evoked Response Auditory

Dealing with expert (human) knowledge consideration, the computer aided medical diagnosis dilemma is one of most interesting, but also one of the most difficult problems. Among difficulties contributing to the challenging nature of this problem, one can mention the need of fine classification. In this paper, we present a new classification approach founded on a tree like neural network based multiple-models structure, able to split a complex problem to a set of simpler sub-problems. This new concept has been used to design a Computer aided medical diagnostic tool that asserts auditory pathologies based on Brain-stem Evoked Response Auditory based biomedical test, which provides an ef-fective measure of the integrity of the auditory pathway.

Reconstruction of 2D parameters using infrared sensors

This paper presents an algorithm used for 2-D parameters estimation, it takes into account the low amount of data provided by dedicated infrared sensors. We use fuzzy modelization to cope with reconstruction uncertainty. The concept of entropy adapted to fuzzy sets is used as a decision criterion to provide a best regularized solution for the ill-posed problem of the reconstruction of 2-D parameters.

On board data fusion and decision system used for obstacle detection: a network and a real time approach

The paper describes a system and an algorithm used for multi-sensorial data fusion. The primary goal is to meet real time constraints with perspectives of low costs products. So the authors have chosen to use binary sensors that supply a relatively low amount of data, which allows the implementation of fast algorithms in order to compute a 2-D representation of a vehicle environment on a one DSP board system. The original parts of this work are located in the definition of system architecture (amount of data dispatching, asynchronism managing and processes localization), and in the implementation of a sub-optimal Kalman like algorithm (with classification rules before decision). the whole system part of a vehicle avoidance demonstrator, and is organized around a PC/AT bus.<<ETX>>

Hybrid HMM/ANN System Using Fuzzy Clustering for Speech and Medical Pattern Recognition

The main goal of this paper is to compare the performance which can be achieved by three different approaches analyzing their applications’ potentiality on real world paradigms. We compare the performance obtained with (1) Discrete Hidden Markov Models (HMM) (2) Hybrid HMM/MLP system using a Multi Layer-Perceptron (MLP) to estimate the HMM emission probabilities and using the K-means algorithm for pattern clustering (3) Hybrid HMM-MLP system using the Fuzzy C-Means (FCM) algorithm for fuzzy pattern clustering.Experimental results on Arabic speech vocabulary and biomedical signals show significant decreases in error rates for the hybrid HMM/MLP system based fuzzy clustering (application of FCM algorithm) in comparison to a baseline system.

Multi-neural Networks Hardware and Software Architecture: Application of the Divide To Simplify Paradigm DTS

We present in this paper the implementation of the data driven method we called DTS (Divide to Simplify), that builds dynamically a Multi-Neural Network Architecture. The Multi-Neural Network architecture, we propose, solves a complex problem by splitting it into several easier problems. We have previously present a software version of the DTS multi-neural network architecture. The main idea of the DTS approach is to use a set of small and specialized mapping neural networks, or Slave Neural Networks (SNN), that are guided by a prototype based neural network, or Master Neural Network (MNN). In this paper, the MNN manages a set of hardware digital neural networks. Learning is performed in few milliseconds. We get a very good rate of classification when using the two spirals problem as a benchmark.

Tree-Like Multiple Neural Network Models Generator with a Complexity Estimation Based Decomposer

In this article we present a self-organizing hybrid modular approach that is aimed at reduction of processing task complexity by decomposition of an initially complex problem into a set of simpler sub-problems. This approach hybridizes artificial neural networks based artificial intelligence and complexity estimation loops in order to reach a higher level intelligent processing capabilities. In consequence, our approach mixtures learning, complexity estimation and specialized data processing modules in order to achieve a higher level self-organizing modular intelligent information processing system. Experimental results validating the presented approach are reported and discussed..

Auto-adaptive neural network tree structure based on complexity estimator

This paper studies the convergence properties of the previously proposed CFA (Clustering for Function Approximation) algorithm and compares its behavior with other input-output clustering techniques also designed for approximation problems. The results obtained show that CFA is able to obtain an initial configuration from which an approximator can improve its performance.