Manuel Samuelides

Institut Supérieur de l'Aéronautique et de l'Espace (ISAE) | ISAE · Mathématiques Appliquées

Introduction Introducing some more notations Linear regression Non-linear regression Kriging interpolation Non-parametric regression and kernel-based methods Support vector regression Model selection Introduction to design of computer experiments (DoCE) Bibliography

Control of amplifier flows poses a great challenge, since the influence of environmental noise sources and measurement contamination is a crucial component in the design of models and the subsequent performance of the controller. A model-based approach that makes a priori assumptions on the noise characteristics often yields unsatisfactory results...

This paper aims to suppress unsteadiness in a convectively unstable flow configuration. System identification methods will be used to design a controller. Unlike classical Galerkin-based methods, no knowledge about the dynamics or about the incoming perturbations is required, as the whole process relies only on data that one could extract from a la...

On présente ici une méthode d?optimisation biniveau de grandes structures de fuselage composite. Ce schéma biniveau est inspiré de la formulation Quasi Separable Decomposition (QSD)récemment développée par Haftka et Watson. Le comportement membrane et hors-plan des stratifiés est représenté au moyen des paramètres de stratification. La boucle d?opt...

An automatic method to combine several local surrogate models is presented. This method is intended to build accurate and smooth approximation of discontinuous functions that are to be used in structural optimization problems. It strongly relies on the Expectation-Maximization (EM) algorithm for Gaussian mixture models (GMM). To the end of regressi...

An optimization framework dedicated to stiffened panel optimization (STIFFOPT) has been implemented and demonstrated on aircraft fuselage covers. The framework incorporates three methodologies, which have been developed with consideration of the design stage: - In rapid sizing approaches the optimization is based on design curves obtained from the...

Le dimensionnement de grandes structures aéronautiques s’appuie sur des heuristiques qui garantissent l’admissibilité de la structure par rapport aux contraintes de tenue mais pas nécessairement l’optimalité en masse. On se propose ici de formaliser le problème d’optimisation de structures en l’incluant dans la catégorie des problèmes d’optimisatio...

In this paper, we propose a Lagrange-Kuhn-Tucker based coordination scheme to deal with level coupling. First, we give a precise definition of several multilevel decomposition schemes, namely the StiffOpt -industrial- approach, and the minMass approach which keeps a mass minimization as objective of local optimisations. Then, we introduce a new dec...

The numerical calculation of the quasi-energy spectrum for a periodically or quasi-periodically kicked rotator supports the evidence in the latter case, of a transition from a regular to an irregular quantum motion.

In this study, we propose a new multilevel optimisation process which responds to these requirements. The original optimization problem is decomposed into a system level and several element levels, aiming at the minimization of element weight under inequality constraints. Bounds of these constraints are given by the system level optimisation. In or...

This special issue of the Journal of Physiology, Paris, is an outcome of NeuroComp'06, the first French conference in Computational Neuroscience. The preparation for this conference, held at Pont-à-Mousson in October 2006, was accompanied by a survey which has resulted in an up-to-date inventory of human resources and labs in France concerned with...

This paper is a review dealing with the study of large size random recurrent neural networks. The connection weights are selected according to a probability law and it is possible to predict the network dynamics at a macroscopic scale using an averaging principle. After a first introductory section, the section 1 reviews the various models from the...

Dans cet article, nous allons exposer deux méthodes de discrimination dont l'expression à partir d'ensembles flous permet d'accéder à de meilleurs résultats en terme de performance pour la première et en terme d'intelligibilité pour l'autre. Toutes deux vont ainsi répondre aux exigences de performance et d'intelligibilité du système opérationnel d'...

This paper presents an overview of some techniques and concepts coming from dynamical system theory and used for the analysis of dynamical neural networks models. In a first section, we describe the dynamics of the neuron, starting from the Hodgkin-Huxley description, which is somehow the canonical description for the "biological neuron". We discus...

In a multilevel optimization frame, the use of surrogate models to approximate optimization constraints allows great time saving. Among available metamodelling techniques we chose to use Neural Networks to perform regression of static mechanical criteria, namely buckling and collapse reserve factors of a stiffened panel, which are constraints of ou...

This article presents a new method for supervised image classification. Given a finite number of image sets, each set corresponding to a place of an environment, we propose a localization strategy, which relies upon supervised classification. For each place, the corresponding landmark is actually a combination of features that have to be detected i...

Recurrent spiking neural networks can provide biologically inspired model of robot controller. We study here the dynamics of large size randomly connected networks thanks to "mean field theory". Mean field theory allows to compute their dynamics under the assumption that the dynamics of individ-ual neuronsare stochastically independent. We restrict...

In the previous chapter, we showed how to use training in order to model controlled dynamical systems, with emphasis on neural modeling. This chapter extends that presentation to the problem of designing a closed-loop control law by training. Nonlinear control has been a growing field during the past twenty years. However, there is no methodology b...

Modeling of controlled dynamical systems or "process identification" is a major application of neural networks. This topic was cursorily addressed in Chap. 2. It is more systematically developed hereafter. Moreover, it is compared to similar statistical methods that are commonly used, especially for linear systems identification. We start with the...

To understand possible strategies of temporal spike coding in the central nervous system, we study functional neuromimetic models of visual processing for static images. We will first present the retinal model which was introduced by Van Rullen and Thorpe and which represents the multiscale contrast values of the image using an orthonormal wavelet...

It is generally assumed that neurons in the central nervous system communicate through temporal ring patterns. As a rst step, we will study the learning of a layer of realistic neurons in the particular case where the relevant messages are formed by temporally correlated patterns, or synre patterns. The model is a layer of Integrate-and-Fire (IF) n...

This paper presents an original application of the Galois lattice theory, the visual landmark selection for topological localization of an autonomous mobile robot, equipped with a color camera. First, visual landmarks have to be selected in order to characterize a structural environment. Second, such landmarks have to be detected and updated for lo...

This paper presents a new decentralized method for selecting visual landmarks in a structured environment. Different images, issued from the different places, are analyzed, and primitives are extracted to determine whether or not features are present in the images. Subsequently, landmarks are selected as a combination of these features with a mathe...

We present here a new methodology to perform active visual localization in the context of autonomous mobile robotics. The robot is endowed with a topological map of its environment. During the learning phase, the robot takes a lot of pictures from the environment; each picture is labelled by its origin place in the topological map. After the learni...

Article dans revue scientifique avec comité de lecture. nationale.

This paper presents a new method for supervised image classification. One or several landmarks are attached to each class, with the intention of characterizing it and discriminating it from the other classes. The different features, deduced from image primitives, and their relationships with the sets of images are structured and organized into a hi...

To explore new coding strategies in the visual system, we defined a general framework of over-complete spike representation by defining lateral interactions [Perrinet et al., 2002]. This algorithm uses arbitrary dictionaries and a greedy matching pursuit to describe static images by spike events as a linear generative model. We investigated a learn...

In order to account for the rapidity of visual processing, we explore visual coding strategies using a one-pass feed-forward spiking neural network. We based our model on the work of Van Rullen and Thorpe Neural Comput. 13 (6) (2001) 1255, which constructs a retinal representation using an orthogonal wavelet transform. This strategy provides a spik...

To explore new coding strategies in the visual system, we defined a general framework of overcomplete representation with spike coding by defining lateral interactions [Perrinet et al., 2002]. This algorithm uses arbitrary dictionaries and a greedy matching pursuit to describe the image by spike events so that images are described by a sparse spike...

In order to account for the rapidity of visual processing, we explore visual coding strategies using a one-pass feed-forward spiking neural network. Following the work of Van Rullen and Thorpe [9], which constructs a spike code for progressive retinal transmission using a wavelet-like transform and rank order coding, we extend this model to arbitra...

We present a new algorithm that extends the Reinforcement Learning framework to Partially Observed Markov Decision Processes (POMDP). The main idea of our method is to build a state extension, called exhaustive observable, which allow us to define a next processus that is Markovian. We bring the proof that solving this new process, to which classic...

We explore visual spike coding strategies in a neural layer in order to build a dynamical model of primary vision. A strictly feed-forward architecture is compared to a strategy accounting for lateral interactions that shows sparse spike coding of the image as is observed in the primary visual areas [1]. This transform is defined over a neural laye...

In order to explore coding strategies in the retina, we use a wavelet-like transform which output is sparse, as is observed in biological retinas [4]. This transform is defined in the context of a one-pass feed-forward spiking neural network, and the output is the list of its neurons' spikes: it is recursively constructed using a greedy matching pu...

We explore visual spike coding strategies in a neural layer in order to build a dynamical model of primary vision. A strictly feed-forward architecture is compared to a strategy accounting for lateral interactions that shows sparse spike coding of the image as is observed in the primary visual areas [Olshausen98]. This transform is defined over a n...

 In this article, we study the asymptotic dynamics of a noisy discrete time neural network, with random asymmetric couplings and thresholds. More precisely, we focus our interest on the limit behaviour of the network when its size grows to infinity with bounded time. In the case of gaussian connection weights, we use the same techniques as Ben Arou...

The French version of this book is out of print. It is available in English: "Neural Networks, methodology and applications", published by Springer. A recent, extended version in French is available: "Apprentissage statistique: réseaux de neurones, cartes topologiques, machines à vecteurs supports" published by Eyrolles.

In this paper, we first present a new mathematical approach, based on large deviation techniques, for the study of a large random recurrent neural network with discrete time dynamics. In particular, we state a mean field property and a law of large numbers, in the most general case of random models with sparse connections and several populations. O...

Based on neurophysiological observations on the behavior of synapses, spike time dependent Hebbian plasticity is a novel extension to the modeling of the Hebb rule. This rule has enormous importance in the learning of spiking neural networks (SNN) but its mechanisms and computational properties are still to be explored.In this article, we present a...

Recent work on biologically motivated networks have shown that the visual system can process a natural scene more quickly by encoding the order of neural firing rather than the frequency of firing. This `order of firing' encoding scheme has led to a rank-based approach which converts activation energy into a time-dependent pulse code. This paper fo...

An image segmentation algorithm, based on Pulse-Coupled Neural Networks, was implemented in silicon. We aimed at simplifying neuron hardware implementation while maintaining segmentation efficiency. Some algorithmic tricks have then been added, improving the results. The main components of the underlying neuron architecture are a single 8 bits regi...

Freeman's investigations on the olfactory bulb of the rabbit showed that its signal dynamics was chaotic, and that recognition of a learned stimulus is linked to a dimension reduction of the dynamics attractor. In this paper we address the question whether this behavior is specific of this particular architecture, or if it is a general property. We...

We propose Hebb-like learning rules to store a static pattern as a dynamical attractor in a neural network with chaotic dynamics. We show that these kind of rules reduces the attractor dimension while learning, until a fixed point is reached. We choose to stop learning when the system is on a strange attractor or limit cycle. We associate this attr...

Recent works have shown that biologically motivated net works of spiking neurons can potentially process information very quickly by encoding information in the latency at which different neurons fire, rather than by using frequency of firing as the code. In this paper, the relevant information is the rank vector of latency order of competing neuro...

Freeman’s investigations on the olfactory bulb of the rabbit showed that its dynamics was chaotic, and that recognition of a learned pattern is linked to a dimension reduction of the dynamics on a much simpler attractor (near limit cycle). We adress here the question wether this behaviour is specific of this particular architecture or if this kind...

Pulsed Coupled Oscillatory Neural Networks are examined for application to image analysis. Adapting to biological constraints, a pulsed coupled network using an Integrate and Fire model with dynamical synapses is examined to perform image segmentation based on synchronisation on the firing time-of neurons which are in the same region. To enhance sy...

Pulsed oscillatory neural networks are examined for application to analysis and segmentation of multispectral imagery from the Satelite Pour l’Observation de la Terre (SPOT). These networks demonstrate a capacity to segment images with better performance against many of the resolution uncertainty effects caused by local area adaptive filtering. To...

An important difference between biological vision systems and their electronic counterparts is the large number of feedback signals controlling each aspect of the image collection process. For every forward path of information in the brain, from sensor to comprehension, there appears to be several neural bundles which send information back to the s...

An important difference between biological vision systems and their electronic counterparts is the large number of feedback signals controlling each aspect of the image collection process. For every forward path of information in the brain, from sensor to comprehension, there appears to be several neural bundles which send information back to the s...

Unlike biological vision, most techniques for computer image processing are not robust over large samples of imagery. Natural systems seem unaffected by variation in local illumination and textures which interfere with conventional analysis. While change detection algorithms have been partially successful, many important tasks like extraction of ro...

The dynamical behaviour of a very general model of neural networks with random asymmetric synaptic weights is investigated in the presence of random thresholds. Using mean-field equations, the bifurcations of the fixed points and the change of regime when varying control parameters are established. Different areas with various regimes are defined i...

Chaos in nervous system is a fascinating but controversial field of investigation. To approach the role of chaos in the real brain, we theoretically and numerically investigate the occurrence of chaos inartificial neural networks. Most of the time, recurrent networks (with feedbacks) are fully connected. This architecture being not biologically pla...

We investigate the dynamical behaviour of neural networks with asymmetric synaptic weights, in the presence of random thresholds. We inspect low gain dynamics before using mean-field equations to study the bifurcations of the fixed points and the change of regime that occurs when varying control parameters. We infer different areas with various reg...

The occurrence of chaos in recurrent neural networks is supposed to depend on the architecture and on the synaptic coupling strength. It is studied here for a randomly diluted architecture. We produce a bifurcation parameter independent of the connectivity that allows a sustained activity and the occurrence of chaos when reaching a critical value....

Summary form only given, as follows. The authors consider a mathematical justification of the offline approximation in continuous-time neural networks. In real-time models, a network behavior is characterized by two distinct dynamics evolving according to different time scales, the weight dynamics which is the `slow' dynamics and the activation dyn...

It is shown that layered feedforward nets are convergent, while nonperiodic oscillations might occur in bidirectional nets. The virtual lateral inhibition phenomenon in feedforward nets is discussed. It is shown that a certain kind of nonlinearity in output functions may be successfully used in realizing winner-take-all networks

Sensitive dependance on initial conditons is studied for a diluted and non symmetric model of neural networks. Two parameters appear to be relevant: the connectivity which allows local difference to spread all over the network and the noise temperature which smooths these difference. Their influence is shown by theoretical estimations in the thermo...

This study provides two results: (1) a theorem that enables a shared resource scheduling problem to be transformed into an unshared resource problem; (2) a competitive activation-based neural network that enables the unshared resource scheduling problem to be solved. These results are used to find an optimal scheduling of antennas of low-level sate...

1. Dans [3], H. Dye dtmontre qu’un groupe a croissance polynomiale qui opere librement sur un espace mesure (X, p) en preservant une mesure finie p definit une relation d’equivalence hyperfinie 99 dont les classes sont les orbites de l’action du groupe. I1 donne d’abord une methode permettant de construire des ensembles mesurables support d’une rel...

We present a discrimation method for seismic events. One event is described by high level features. Since these variables are both quantitative and qualitative, we develop a processing line, on the cross-road of statistics (”Mixtures of Experts”) and Artificial Intelligence (”Fuzzy Inference System”). It can be viewed as an original extension of Ra...

This paper presents a strategy to build more accurate databases for response surface approximation by neural networks. A sequential enrichment of the database based on cross-validation and bootstrap techniques is proposed to obtain adapted response surfaces for optimization. The proposed algorithm is evaluated on a two phase flow configuration.

Ce papier présente une nouvelle technique pour la localisation d'un robot mobile autonome dans un environnement structuré. La localisation est topologique et se base sur les amers visuels. Ces amers sont des combinaisons de caractéristiques visuelles sélectionnées à l'aide d'un formalisme mathématique appelé treillis de Galois, ou treillis de conce...

Ce papier présente une nouvelle méthode pour la classification supervisée d'ensembles d'images. A chacun de ces ensembles est attaché, pour le caractériser et le distinguer des autres, une entité appelée amer. Différentes caractéristiques, déduites de primitives visuelles, sont extraites de chaque image, et la relation qui lie certaines images à ce...

L'apprentissage statistique permet la mise au point de modèles de données et de processus lorsque la formalisation de règles explicites serait impossible : reconnaissance de formes ou de signaux, prévision, fouille de données, prise de décision en environnement complexe et évolutif. Ses applications sont multiples dans le monde de la production ind...

Dans ce chapitre, nous exposons systématiquement et en général les méthodes de construction des modèles réduits dits "non physiques" ou "méta-modèles" (en anglais surrogate models) car construits à partir de données calculées sur des modèles physiques fins ou détaillés). Nous commençons par exposer des rappels sur les méthodes de régression (sectio...

Mémoire DES : U.E.R.: 02 / Paris I ; 1974 : sess. d'oct. Multigraphié.