Adel M. Alimi

Bio: Adel M. Alimi is an academic researcher from University of Sfax. The author has contributed to research in topics: Fuzzy logic & Artificial neural network. The author has an hindex of 40, co-authored 716 publications receiving 9168 citations. Previous affiliations of Adel M. Alimi include Eni & École Polytechnique de Montréal.

Speed/accuracy trade-offs in target-directed movements.

Abstract: This target article presents a critical survey of the scientific literature dealing with the speed/accuracy trade-offs in rapid-aimed movements. It highlights the numerous mathematical and theoretical interpretations that have been proposed in recent decades. Although the variety of points of view reflects the richness of the field and the high degree of interest that such basic phenomena attract in the understanding of human movements, it calls into question the ability of 'many models to explain the basic observations consistently reported in the field. This target article summarizes the kinematic theory of rapid human movements, proposed recently by R. Plamondon (1993b; 1993c; 1995a; 1995b), and analyzes its predictions in the context of speed/accuracy trade-offs. Data from human movement literature are reanalyzed and reinterpreted in the context of the new theory. It is shown that the various aspects of speed/accuracy trade- offs can be taken into account by considering the asymptotic behavior of a large number of coupled linear systems, from which a delta- lognormal law can be derived to describe the velocity profile of an end-effector driven by a neuromuscular synergy. This law not only describes velocity profiles almost perfectly, it also predicts the kinematic properties of simple rapid movements and provides a consistent framework for the analysis of different types of speed/accuracy trade-offs using a quadratic (or power) law that emerges from the model.

PSO-based analysis of Echo State Network parameters for time series forecasting

Abstract: Graphical abstractDisplay Omitted HighlightsEcho State Network (ESN) is an interesting tool for dealing with time series forecasting problems.The learning performance of ESN can be affected because of the random setting of some untrained weights.PSO is introduced to ESN as a pre-training tool to optimize the untrained weights.The networks weights become suitable to the targeted application.The accuracy of ESN is considerably improved after PSO pre-training. Echo State Networks, ESNs, are standardly composed of additive units undergoing sigmoid function activation. They consist of a randomly recurrent neuronal infra-structure called reservoir. Coming up with a good reservoir depends mainly on picking up the right parameters for the network initialization. Human expertise as well as repeatedly tests may sometimes provide acceptable parameters. Nevertheless, they are non-guaranteed. On the other hand, optimization techniques based on evolutionary learning have proven their strong effectiveness in unscrambling optimal solutions in complex spaces. Particle swarm optimization (PSO) is one of the most popular continuous evolutionary algorithms. Throughout this paper, a PSO algorithm is associated to ESN to pre-train some fixed weights values within the network. Once the network's initial parameters are set, some untrained weights are selected for optimization. The new weights, already optimized, are re-squirted to the network which launches its normal training process. The performances of the network are a subject of the error and the time processing evaluation metrics. The testing results after PSO pre-training are compared to those of ESN without optimization and other existent approaches. The conceived approach is tested for time series prediction purpose on a set of benchmarks and real-life datasets. Experimental results show obvious enhancement of ESN learning results.

Multi-objective Feature Selection with NSGA II

Abstract: This paper deals with the multi-objective definition of the feature selection problem for different pattern recognition domains. We use NSGA II the latest multi-objective algorithm developed for resolving problems of multi-objective aspects with more accuracy and a high convergence speed. We define the feature selection as a problem including two competing objectives and we try to find a set of optimal solutions so called Pareto-optimal solutions instead of a single optimal solution. The two competing objectives are the minimization of both the number of used features and the classification error using 1-NN classifier. We apply our method to five databases selected from the UCI repository and we report the results on these databases. We present the convergence of the NSGA II on different problems and discuss the behavior of NSGA II on these different contexts.

The Affective Tutoring System

Abstract: Affective Computing is a new Artificial Intelligence area that deals with the possibility of making computers able to recognize human emotions in different ways. This paper represents a study about the integration of this new area in the intelligent tutoring system. We argue that socially appropriate affective behaviors would provide a new dimension for collaborative learning systems. The main goal is to analyses learner facial expressions and show how Affective Computing could contribute for this interaction, being part of the complete student tracking (traceability) to monitor student behaviors during learning sessions.

Finite-time and fixed-time synchronization of a class of inertial neural networks with multi-proportional delays and its application to secure communication

Abstract: Proportional delay, which is different from time-varying delays and distributed delay, is a kind of unbounded delay. The proportional delay system as an important mathematical model often rises in some various fields such as control theory, physics and biology systems. This paper is concerned with the finite-time and the fixed-time synchronization problem for a class of inertial neural networks with multi-proportional delays. First, by constructing a proper variable substitution, the original inertial neural networks with multi-proportional delays can be rewritten as a first-order differential system. Second, by constructing Lyapunov functionals and by using analytical techniques, and together with novel control algorithms, some new and effective criteria are established to achieve finite-time and fixed-time synchronization of the master/slave of addressed systems. Finally, several examples and their simulations are given to illustrate the effectiveness of the proposed method. Furthermore, a secure communication synchronization problem is presented to illustrate the effectiveness of the obtained results.

Statistical method for testing the neutral mutation hypothesis by DNA polymorphism.

Abstract: The relationship between the two estimates of genetic variation at the DNA level, namely the number of segregating sites and the average number of nucleotide differences estimated from pairwise comparison, is investigated. It is found that the correlation between these two estimates is large when the sample size is small, and decreases slowly as the sample size increases. Using the relationship obtained, a statistical method for testing the neutral mutation hypothesis is developed. This method needs only the data of DNA polymorphism, namely the genetic variation within population at the DNA level. A simple method of computer simulation, that was used in order to obtain the distribution of a new statistic developed, is also presented. Applying this statistical method to the five regions of DNA sequences in Drosophila melanogaster, it is found that large insertion/deletion (greater than 100 bp) is deleterious. It is suggested that the natural selection against large insertion/deletion is so weak that a large amount of variation is maintained in a population.

The Self-Organizing Map

Abstract: An overview of the self-organizing map algorithm, on which the papers in this issue are based, is presented in this article.

Supervised Sequence Labelling with Recurrent Neural Networks

Abstract: Recurrent neural networks are powerful sequence learners. They are able to incorporate context information in a flexible way, and are robust to localised distortions of the input data. These properties make them well suited to sequence labelling, where input sequences are transcribed with streams of labels. The aim of this thesis is to advance the state-of-the-art in supervised sequence labelling with recurrent networks. Its two main contributions are (1) a new type of output layer that allows recurrent networks to be trained directly for sequence labelling tasks where the alignment between the inputs and the labels is unknown, and (2) an extension of the long short-term memory network architecture to multidimensional data, such as images and video sequences.