Dissertations / Theses on the topic 'Réseaux de neurones non supervisé'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Réseaux de neurones non supervisé.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
Galtier, Mathieu. "Une approche mathématique de l'apprentissage non-supervisé dans les réseaux de neurones récurrents." Phd thesis, École Nationale Supérieure des Mines de Paris, 2011. http://pastel.archives-ouvertes.fr/pastel-00667368.
Full textYin, Hao. "Étude des réseaux de neurones en mode non supervisé : application à la reconnaissance des formes." Compiègne, 1992. http://www.theses.fr/1992COMPD524.
Full textCherif, Aymen. "Réseaux de neurones, SVM et approches locales pour la prévision de séries temporelles." Thesis, Tours, 2013. http://www.theses.fr/2013TOUR4003/document.
Full textTime series forecasting is a widely discussed issue for many years. Researchers from various disciplines have addressed it in several application areas : finance, medical, transportation, etc. In this thesis, we focused on machine learning methods : neural networks and SVM. We have also been interested in the meta-methods to push up the predictor performances, and more specifically the local models. In a divide and conquer strategy, the local models perform a clustering over the data sets before different predictors are affected into each obtained subset. We present in this thesis a new algorithm for recurrent neural networks to use them as local predictors. We also propose two novel clustering techniques suitable for local models. The first is based on Kohonen maps, and the second is based on binary trees
Manenti, Céline. "Découverte d'unités linguistiques à l'aide de méthodes d'apprentissage non supervisé." Thesis, Toulouse 3, 2019. http://www.theses.fr/2019TOU30074.
Full textThe discovery of elementary linguistic units (phonemes, words) only from sound recordings is an unresolved problem that arouses a strong interest from the community of automatic speech processing, as evidenced by the many recent contributions of the state of the art. During this thesis, we focused on using neural networks to answer the problem. We approached the problem using neural networks in a supervised, poorly supervised and multilingual manner. We have developed automatic phoneme segmentation and phonetic classification tools based on convolutional neural networks. The automatic segmentation tool obtained 79% F-measure on the BUCKEYE conversational speech corpus. This result is similar to a human annotator according to the inter-annotator agreement provided by the creators of the corpus. In addition, it does not need a lot of data (about ten minutes per speaker and 5 different speakers) to be effective. In addition, it is portable to other languages (especially for poorly endowed languages such as xitsonga). The phonetic classification system makes it possible to set the various parameters and hyperparameters that are useful for an unsupervised scenario. In the unsupervised context, the neural networks (Auto-Encoders) allowed us to generate new parametric representations, concentrating the information of the input frame and its neighboring frames. We studied their utility for audio compression from the raw signal, for which they were effective (low RMS, even at 99% compression). We also carried out an innovative pre-study on a different use of neural networks, to generate vectors of parameters not from the outputs of the layers but from the values of the weights of the layers. These parameters are designed to mimic Linear Predictive Coefficients (LPC). In the context of the unsupervised discovery of phoneme-like units (called pseudo-phones in this memory) and the generation of new phonetically discriminative parametric representations, we have coupled a neural network with a clustering tool (k-means ). The iterative alternation of these two tools allowed the generation of phonetically discriminating parameters for the same speaker: low rates of intra-speaker ABx error of 7.3% for English, 8.5% for French and 8 , 4% for Mandarin were obtained. These results allow an absolute gain of about 4% compared to the baseline (conventional parameters MFCC) and are close to the best current approaches (1% more than the winner of the Zero Resource Speech Challenge 2017). The inter-speaker results vary between 12% and 15% depending on the language, compared to 21% to 25% for MFCCs
Delsert, Stéphane. "Classification interactive non supervisée de données multidimensionnelles par réseaux de neurones à apprentissage cométitif." Lille 1, 1996. https://pepite-depot.univ-lille.fr/LIBRE/Th_Num/1996/50376-1996-214.pdf.
Full textBernert, Marie. "Développement d'un réseau de neurones STDP pour le tri en ligne et non-supervisé de potentiels d'action." Thesis, Université Grenoble Alpes (ComUE), 2019. http://www.theses.fr/2019GREAS001.
Full textPattern recognition is a fundamental task for living beings and is perform very efficiently by the brain. Artificial deep neural networks are making quick progress in reproducing these performance and have many applications such as image recognition or natural language processing. However, they require extensive training on large datasets and heavy computations. A promising alternative are spiking neural networks, which closely mimic what happens in the brain, with spiking neurons and spike-timing dependent plasticity (STDP). They are able to perform unsupervised learning and have been used for visual or auditory pattern recognition. However, for now applications using STDP networks lag far behind classical deep learning. Developing new applications for this kind of networks is all the more at stake that they could be implemented in low power neuromorphic hardware that currently undergoes important developments, in particular with analog miniaturized memristive devices able to mimic synaptic plasticity. In this work, we chose to develop an STDP neural network to perform a specific task: spike-sorting, which is a crucial problem in neuroscience. Brain implants based on microelectrode arrays are able to record the activity of individual neurons, appearing in the recorded signal as peak potential variations called action potentials. However, several neurons can be recorded by the same electrode. The goal of spike-sorting is to extract and separate the activity of different neural cells from a common extracellular recording taking advantage of the fact that the shape of an action potential on an electrode depends on the neuron it stems from. Thus spike-sorting can be seen as an unsupervised pattern recognition task where the goal is to detect and classify different waveforms. Most classical spike-sorting approaches use three separated steps: detecting all action potentials in the signal, extract features characterizing their shapes, and separating these features into clusters that should correspond to different neural cells. Though online methods exists, most widespread spike-sorting methods are offline or require an offline preprocessing step, which is not compatible with online application such as Brain-computer interfaces (BCI). Moreover, the development of always larger microelectrode arrays creates a need for fully automatic and computationally efficient algorithms. Using an STDP network brings a new approach to meet these requirements. We designed a network that take the electrode signal as an input, and output spikes that correspond to the spiking activity of the recorded neural cells. It is organized into several layers, designed to achieve different processing steps, connected in feedforward way. The first layer, composed of neurons acting as sensory neurons, convert the input signal into spike train. The following layers are able to learn patterns from the previous layer thanks to STDP rules. Each layer implement different mechanisms that improve their performance, such as resource-dependent STDP, intrinsic plasticity, plasticity triggered by inhibition, or neuron models having rebound spiking properties. An attention mechanism has been implemented to make the network sensitive only to part of the signal containing action potentials. This network was first designed to process data from a single electrode, and then adapted to process data from multiple electrodes. It has been tested on simulated data, which allowed to compare the network output to the known ground truth, and also on real extracellular recordings associated with intracellular recordings that give an incomplete ground truth. Different versions of the network were evaluated and compared to other spike-sorting algorithms, and found to give very satisfying results. Following these software simulations, we initiated an FPGA implementation of the method, which constitutes a first step toward embedded neuromorphic implementation
Buhot, Arnaud. "Etude de propriétés d'apprentissage supervisé et non supervisé par des méthodes de Physique Statistique." Phd thesis, Université Joseph Fourier (Grenoble), 1999. http://tel.archives-ouvertes.fr/tel-00001642.
Full textSchutz, Georges. "Adaptations et applications de modèles mixtes de réseaux de neurones à un processus industriel." Phd thesis, Université Henri Poincaré - Nancy I, 2006. http://tel.archives-ouvertes.fr/tel-00115770.
Full textartificiels pour améliorer le contrôle de processus industriels
complexes, caractérisés en particulier par leur aspect temporel.
Les motivations principales pour traiter des séries temporelles
sont la réduction du volume de données, l'indexation pour la
recherche de similarités, la localisation de séquences,
l'extraction de connaissances (data mining) ou encore la
prédiction.
Le processus industriel choisi est un four à arc
électrique pour la production d'acier liquide au Luxembourg. Notre
approche est un concept de contrôle prédictif et se base sur des
méthodes d'apprentissage non-supervisé dans le but d'une
extraction de connaissances.
Notre méthode de codage se base sur
des formes primitives qui composent les signaux. Ces formes,
composant un alphabet de codage, sont extraites par une méthode
non-supervisée, les cartes auto-organisatrices de Kohonen (SOM).
Une méthode de validation des alphabets de codage accompagne
l'approche.
Un sujet important abordé durant ces recherches est
la similarité de séries temporelles. La méthode proposée est
non-supervisée et intègre la capacité de traiter des séquences de
tailles variées.
Lefort, Mathieu. "Apprentissage spatial de corrélations multimodales par des mécanismes d'inspiration corticale." Phd thesis, Université Nancy II, 2012. http://tel.archives-ouvertes.fr/tel-00756687.
Full textGermain, Mathieu. "L’estimation de distribution à l'aide d'un autoencodeur." Mémoire, Université de Sherbrooke, 2015. http://hdl.handle.net/11143/6910.
Full textMaurel, Denis. "Contributions aux communications inter-vues pour l'apprentissage collaboratif." Electronic Thesis or Diss., Sorbonne université, 2018. http://www.theses.fr/2018SORUS489.
Full textThis thesis presents several methods to optimize and improve inter-views communications in a collaborative learning context: The first contribution is about the improvement of communications for Collaborative Clustering using a learning method making it possible for a view to weight the information supplied by the external views. This methods is based on the resolution of a problem made of the Collaborative Clustering criterion with two constraints of the weighting coefficients. A second contribution consists in the definition of an incremental learning method of Self-Organizing Maps, followed by its adaptation to Collaborative Clustering. This method makes it possible to adapt the results obtained using Collaborative Clustering in case of a potential evolution in data distribution through time. The second axis consists in the definition of a new paradigm, called Collaborative Reconstruction. In this paradigm, several views collaborate to reconstruct local missing data. This method is based on neural networks linking external data and local data. The combination of the external informations is guaranteed by a weighting method favoring the best reconstructed features for each external view
Le, Lan Gaël. "Analyse en locuteurs de collections de documents multimédia." Thesis, Le Mans, 2017. http://www.theses.fr/2017LEMA1020/document.
Full textThe task of speaker diarization and linking aims at answering the question "who speaks and when?" in a collection of multimedia recordings. It is an essential step to index audiovisual contents. The task of speaker diarization and linking firstly consists in segmenting each recording in terms of speakers, before linking them across the collection. Aim is, to identify each speaker with a unique anonymous label, even for speakers appearing in multiple recordings, without any knowledge of their identity or number. The challenge of the cross-recording linking is the modeling of the within-speaker/across-recording variability: depending on the recording, a same speaker can appear in multiple acoustic conditions (in a studio, in the street...). The thesis proposes two methods to overcome this issue. Firstly, a novel neural variability compensation method is proposed, using the triplet-loss paradigm for training. Secondly, an iterative unsupervised domain adaptation process is presented, in which the system exploits the information (even inaccurate) about the data it processes, to enhance its performances on the target acoustic domain. Moreover, novel ways of analyzing the results in terms of speaker are explored, to understand the actual performance of a diarization and linking system, beyond the well-known Diarization Error Rate (DER). Systems and methods are evaluated on two TV shows of about 40 episodes, using either a global, or longitudinal linking architecture, and state of the art speaker modeling (i-vector)
Oquab, Maxime. "Convolutional neural networks : towards less supervision for visual recognition." Thesis, Paris Sciences et Lettres (ComUE), 2018. http://www.theses.fr/2018PSLEE061.
Full textConvolutional Neural Networks are flexible learning algorithms for computer vision that scale particularly well with the amount of data that is provided for training them. Although these methods had successful applications already in the ’90s, they were not used in visual recognition pipelines because of their lesser performance on realistic natural images. It is only after the amount of data and the computational power both reached a critical point that these algorithms revealed their potential during the ImageNet challenge of 2012, leading to a paradigm shift in visual recogntion. The first contribution of this thesis is a transfer learning setup with a Convolutional Neural Network for image classification. Using a pre-training procedure, we show that image representations learned in a network generalize to other recognition tasks, and their performance scales up with the amount of data used in pre-training. The second contribution of this thesis is a weakly supervised setup for image classification that can predict the location of objects in complex cluttered scenes, based on a dataset indicating only with the presence or absence of objects in training images. The third contribution of this thesis aims at finding possible paths for progress in unsupervised learning with neural networks. We study the recent trend of Generative Adversarial Networks and propose two-sample tests for evaluating models. We investigate possible links with concepts related to causality, and propose a two-sample test method for the task of causal discovery. Finally, building on a recent connection with optimal transport, we investigate what these generative algorithms are learning from unlabeled data
Falez, Pierre. "Improving spiking neural networks trained with spike timing dependent plasticity for image recognition." Thesis, Lille 1, 2019. http://www.theses.fr/2019LIL1I101.
Full textComputer vision is a strategic field, in consequence of its great number of potential applications which could have a high impact on society. This area has quickly improved over the last decades, especially thanks to the advances of artificial intelligence and more particularly thanks to the accession of deep learning. Nevertheless, these methods present two main drawbacks in contrast with biological brains: they are extremely energy intensive and they need large labeled training sets. Spiking neural networks are alternative models offering an answer to the energy consumption issue. One attribute of these models is that they can be implemented very efficiently on hardware, in order to build ultra low-power architectures. In return, these models impose certain limitations, such as the use of only local memory and computations. It prevents the use of traditional learning methods, for example the gradient back-propagation. STDP is a learning rule, observed in biology, which can be used in spiking neural networks. This rule reinforces the synapses in which local correlations of spike timing are detected. It also weakens the other synapses. The fact that it is local and unsupervised makes it possible to abide by the constraints of neuromorphic architectures, which means it can be implemented efficiently, but it also provides a solution to the data set labeling issue. However, spiking neural networks trained with the STDP rule are affected by lower performances in comparison to those following a deep learning process. The literature about STDP still uses simple data but the behavior of this rule has seldom been used with more complex data, such as sets made of a large variety of real-world images.The aim of this manuscript is to study the behavior of these spiking models, trained through the STDP rule, on image classification tasks. The main goal is to improve the performances of these models, while respecting as much as possible the constraints of neuromorphic architectures. The first contribution focuses on the software simulations of spiking neural networks. Hardware implementation being a long and costly process, using simulation is a good alternative in order to study more quickly the behavior of different models. Then, the contributions focus on the establishment of multi-layered spiking networks; networks made of several layers, such as those in deep learning methods, allow to process more complex data. One of the chapters revolves around the matter of frequency loss seen in several spiking neural networks. This issue prevents the stacking of multiple spiking layers. The center point then switches to a study of STDP behavior on more complex data, especially colored real-world image. Multiple measurements are used, such as the coherence of filters or the sparsity of activations, to better understand the reasons for the performance gap between STDP and the more traditional methods. Lastly, the manuscript describes the making of multi-layered networks. To this end, a new threshold adaptation mechanism is introduced, along with a multi-layer training protocol. It is proven that such networks can improve the state-of-the-art for STDP
Labonne, Maxime. "Anomaly-based network intrusion detection using machine learning." Electronic Thesis or Diss., Institut polytechnique de Paris, 2020. http://www.theses.fr/2020IPPAS011.
Full textIn recent years, hacking has become an industry unto itself, increasing the number and diversity of cyber attacks. Threats on computer networks range from malware to denial of service attacks, phishing and social engineering. An effective cyber security plan can no longer rely solely on antiviruses and firewalls to counter these threats: it must include several layers of defence. Network-based Intrusion Detection Systems (IDSs) are a complementary means of enhancing security, with the ability to monitor packets from OSI layer 2 (Data link) to layer 7 (Application). Intrusion detection techniques are traditionally divided into two categories: signatured-based (or misuse) detection and anomaly detection. Most IDSs in use today rely on signature-based detection; however, they can only detect known attacks. IDSs using anomaly detection are able to detect unknown attacks, but are unfortunately less accurate, which generates a large number of false alarms. In this context, the creation of precise anomaly-based IDS is of great value in order to be able to identify attacks that are still unknown.In this thesis, machine learning models are studied to create IDSs that can be deployed in real computer networks. Firstly, a three-step optimization method is proposed to improve the quality of detection: 1/ data augmentation to rebalance the dataset, 2/ parameters optimization to improve the model performance and 3/ ensemble learning to combine the results of the best models. Flows detected as attacks can be analyzed to generate signatures to feed signature-based IDS databases. However, this method has the disadvantage of requiring labelled datasets, which are rarely available in real-life situations. Transfer learning is therefore studied in order to train machine learning models on large labeled datasets, then finetune them on benign traffic of the network to be monitored. This method also has flaws since the models learn from already known attacks, and therefore do not actually perform anomaly detection. Thus, a new solution based on unsupervised learning is proposed. It uses network protocol header analysis to model normal traffic behavior. Anomalies detected are then aggregated into attacks or ignored when isolated. Finally, the detection of network congestion is studied. The bandwidth utilization between different links is predicted in order to correct issues before they occur
Mehr, Éloi. "Unsupervised Learning of 3D Shape Spaces for 3D Modeling." Electronic Thesis or Diss., Sorbonne université, 2019. http://www.theses.fr/2019SORUS566.
Full textEven though 3D data is becoming increasingly more popular, especially with the democratization of virtual and augmented experiences, it remains very difficult to manipulate a 3D shape, even for designers or experts. Given a database containing 3D instances of one or several categories of objects, we want to learn the manifold of plausible shapes in order to develop new intelligent 3D modeling and editing tools. However, this manifold is often much more complex compared to the 2D domain. Indeed, 3D surfaces can be represented using various embeddings, and may also exhibit different alignments and topologies. In this thesis we study the manifold of plausible shapes in the light of the aforementioned challenges, by deepening three different points of view. First of all, we consider the manifold as a quotient space, in order to learn the shapes’ intrinsic geometry from a dataset where the 3D models are not co-aligned. Then, we assume that the manifold is disconnected, which leads to a new deep learning model that is able to automatically cluster and learn the shapes according to their typology. Finally, we study the conversion of an unstructured 3D input to an exact geometry, represented as a structured tree of continuous solid primitives
Hammadi, Youssef. "Réduction d'un modèle 0D instationnaire et non-linéaire de thermique habitacle pour l’optimisation énergétique des véhicules automobiles." Thesis, Université Paris sciences et lettres, 2020. http://www.theses.fr/2020UPSLM027.
Full textThe use of automotive air conditioning leads to a fuel overconsumption. To reduce this overconsumption, we can either work upstream on the technical definitions of the cabin and the HVAC system or optimize control strategies. In both cases, it is essential to build a cabin thermal model that well balances accuracy and complexity. This is the topic of this PhD thesis driven by Renault Group. First, a model reduction methodology is used to build a 0D model starting from a 3D finite element cabin thermal model. This 0D model is based on mass and energy balances on the different cabin walls and air zones. It consists of a nonlinear differential algebraic equations system which can be reinterpreted as a Bond Graph. In addition, the 0D model is based on a weak coupling between the thermal equations and the fluid mechanics ones resulting from CFD calculations (internal airflow and external aerodynamics). Secondly, we apply a machine learning method to the data generated by the 0D model in order to build a reduced 0D model. A design of experiment is considered at this stage. Due to the nonlinearity of the heat exchanges, we have developed an approach which is inspired by the Gappy POD and EIM methods. We use a multiphysics reduced basis that takes several contributions into account (temperatures, enthalpies, heat fluxes and humidities). The resulting reduced model is a hybrid model that couples some of the original physical equations to an artificial neural network. The reduction methodology has been validated on Renault vehicles. The reduced order models have been integrated into a vehicle system-level energetic simulation platform (GREEN) which models different thermics (engine, transmission, cooling system, battery, HVAC, refrigerant circuit, underhood) in order to perform thermal management studies which are of particular importance for electric and hybrid vehicles. The reduced order models have been validated on several scenarios (temperature control for thermal comfort, driving cycles, HVAC coupling) and have achieved CPU gains of up to 99% with average errors of 0.5 °C on temperatures and 0.6% on relative humidities
Al, Chami Zahi. "Estimation de la qualité des données multimedia en temps réel." Thesis, Pau, 2021. http://www.theses.fr/2021PAUU3066.
Full textOver the past decade, data providers have been generating and streaming a large amount of data, including images, videos, audio, etc. In this thesis, we will be focusing on processing images since they are the most commonly shared between the users on the global inter-network. In particular, treating images containing faces has received great attention due to its numerous applications, such as entertainment and social media apps. However, several challenges could arise during the processing and transmission phase: firstly, the enormous number of images shared and produced at a rapid pace requires a significant amount of time to be processed and delivered; secondly, images are subject to a wide range of distortions during the processing, transmission, or combination of many factors that could damage the images’content. Two main contributions are developed. First, we introduce a Full-Reference Image Quality Assessment Framework in Real-Time, capable of:1) preserving the images’content by ensuring that some useful visual information can still be extracted from the output, and 2) providing a way to process the images in real-time in order to cope with the huge amount of images that are being received at a rapid pace. The framework described here is limited to processing those images that have access to their reference version (a.k.a Full-Reference). Secondly, we present a No-Reference Image Quality Assessment Framework in Real-Time. It has the following abilities: a) assessing the distorted image without having its distortion-free image, b) preserving the most useful visual information in the images before publishing, and c) processing the images in real-time, even though the No-Reference image quality assessment models are considered very complex. Our framework offers several advantages over the existing approaches, in particular: i. it locates the distortion in an image in order to directly assess the distorted parts instead of processing the whole image, ii. it has an acceptable trade-off between quality prediction accuracy and execution latency, andiii. it could be used in several applications, especially these that work in real-time. The architecture of each framework is presented in the chapters while detailing the modules and components of the framework. Then, a number of simulations are made to show the effectiveness of our approaches to solve our challenges in relation to the existing approaches
Wynen, Daan. "Une représentation archétypale de style artistique : résumer et manipuler des stylesartistiques d'une façon interprétable." Thesis, Université Grenoble Alpes, 2020. http://www.theses.fr/2020GRALM066.
Full textIn this thesis we study the representations used to describe and manipulate artistic style of visual arts.In the neural style transfer literature and related strains of research, different representations have been proposed, but in recent years the by far dominant representations of artistic style in the computer vision community have been those learned by deep neural networks, trained on natural images.We build on these representations with the dual goal of summarizing the artistic styles present in large collections of digitized artworks, as well as manipulating the styles of images both natural and artistic.To this end, we propose a concise and intuitive representation based on archetypal analysis, a classic unsupervised learning method with properties that make it especially suitable for the task. We demonstrate how this archetypal representation of style can be used to discover and describe, in an interpretable way, which styles are present in a large collection.This enables the exploration of styles present in a collection from different angles; different ways of visualizing the information allow for different questions to be asked.These can be about a style that was identified across artworks, about the style of a particular artwork, or more broadly about how the styles that were identified relate to one another.We apply our analysis to a collection of artworks obtained from WikiArt, an online collection effort of visual arts driven by volunteers. This dataset also includes metadata such as artist identies, genre, and style of the artworks. We use this metadata for further analysis of the archetypal style representation along biographic lines of artists and with an eye on the relationships within groups of artists
Peltier, Marie-Agnès. "Un système adaptatif de diagnostic d'évolution basé sur la reconnaissance des formes floues : application au diagnostic du comportement d'un conducteur automobile." Compiègne, 1993. http://www.theses.fr/1993COMPD634.
Full textBelharbi, Soufiane. "Neural networks regularization through representation learning." Thesis, Normandie, 2018. http://www.theses.fr/2018NORMIR10/document.
Full textNeural network models and deep models are one of the leading and state of the art models in machine learning. They have been applied in many different domains. Most successful deep neural models are the ones with many layers which highly increases their number of parameters. Training such models requires a large number of training samples which is not always available. One of the fundamental issues in neural networks is overfitting which is the issue tackled in this thesis. Such problem often occurs when the training of large models is performed using few training samples. Many approaches have been proposed to prevent the network from overfitting and improve its generalization performance such as data augmentation, early stopping, parameters sharing, unsupervised learning, dropout, batch normalization, etc. In this thesis, we tackle the neural network overfitting issue from a representation learning perspective by considering the situation where few training samples are available which is the case of many real world applications. We propose three contributions. The first one presented in chapter 2 is dedicated to dealing with structured output problems to perform multivariate regression when the output variable y contains structural dependencies between its components. Our proposal aims mainly at exploiting these dependencies by learning them in an unsupervised way. Validated on a facial landmark detection problem, learning the structure of the output data has shown to improve the network generalization and speedup its training. The second contribution described in chapter 3 deals with the classification task where we propose to exploit prior knowledge about the internal representation of the hidden layers in neural networks. This prior is based on the idea that samples within the same class should have the same internal representation. We formulate this prior as a penalty that we add to the training cost to be minimized. Empirical experiments over MNIST and its variants showed an improvement of the network generalization when using only few training samples. Our last contribution presented in chapter 4 showed the interest of transfer learning in applications where only few samples are available. The idea consists in re-using the filters of pre-trained convolutional networks that have been trained on large datasets such as ImageNet. Such pre-trained filters are plugged into a new convolutional network with new dense layers. Then, the whole network is trained over a new task. In this contribution, we provide an automatic system based on such learning scheme with an application to medical domain. In this application, the task consists in localizing the third lumbar vertebra in a 3D CT scan. A pre-processing of the 3D CT scan to obtain a 2D representation and a post-processing to refine the decision are included in the proposed system. This work has been done in collaboration with the clinic "Rouen Henri Becquerel Center" who provided us with data
Khacef, Lyes. "Exploration du calcul bio-inspiré avec des architectures neuromorphiques auto-organisées." Thesis, Université Côte d'Azur, 2020. http://www.theses.fr/2020COAZ4085.
Full textThe brain's cortical plasticity is one of the main features that enable our capability to learn and adapt in our environment. Indeed, the cerebral cortex has the ability to self-organize itself through two distinct forms of plasticity: the structural plasticity that creates (sprouting) or cuts (pruning) synaptic connections between neurons, and the synaptic plasticity that modifies the synaptic connections strength. These mechanisms are very likely at the basis of an extremely interesting characteristic of the human brain development: the multimodal association. In spite of the diversity of the sensory modalities, like sight, sound and touch, the brain arrives at the same concepts. Moreover, biological observations show that one modality can activate the internal representation of another modality when both are correlated. To model such a behavior, Edelman and Damasio proposed respectively the Reentry and the Convergence Divergence Zone frameworks where bi-directional neural communications can lead to both multimodal fusion (convergence) and inter-modal activation (divergence). Nevertheless, these theoretical frameworks do not provide a computational model at the neuron level.The objective of this thesis is first to explore the foundations of brain-inspired self-organization in terms of (1) multimodal unsupervised learning, (2) massively parallel, distributed and local computing, and (3) extremely energy-efficient processing. Based on these guidelines and a review of the neural models in the literature, we choose the Self-Organizing Map (SOM) proposed by Kohonen as the main component of our system. We introduce the Iterative Grid, a fully distributed architecture with local connectivity amongst hardware neurons which enables cellular computing in the SOM, and thus a scalable system is terms of processing time and connectivity complexity.Then, we assess the performance of the SOM in the problem of post-labeled unsupervised learning: no label is available during training, then very few labels are available for naming the SOM neurons. We propose and compare different labeling methods so that we minimize the number of labels while keeping the best accuracy. We compare our performance to a different approach using Spiking Neural Networks (SNNs) with Spike Timing Dependant Plasticity (STDP) learning.Next, we propose to improve the SOM performance by using extracted features instead of raw data. We conduct a comparative study on the SOM classification accuracy with unsupervised feature extraction from the MNIST dataset using two different approaches: a machine learning approach with Sparse Convolutional Auto-Encoders using gradient-based learning, and a neuroscience approach with SNNs using STDP learning.To prove the SOM ability to handle more complex datasets, we use transfer learning in the mini-ImageNet few shot classification benchmark to exploit a Wide Residual Network backbone trained on a base dataset as a feature extractor, then we use the SOM to classify the obtained features from the target dataset.Finally, we move into the multimodal association mechanism. We build the Reentrant SOM (ReSOM), a brain-inspired neural system based on the Reentry principles using SOMs and Hebbian-like learning. We propose and compare different computational methods for multimodal unsupervised learning and inference, then quantify the gain of both convergence and divergence mechanisms on three multimodal datasets. The divergence mechanism is used to label one modality based on the other, while the convergence mechanism is used to improve the overall accuracy of the system. We compare our results to SNNs with STDP learning and different fusion strategies, then we show the gain of the so-called hardware plasticity induced by our model, where the system's topology is not fixed by the user but learned along the system's experience through self-organization
Kassab, Randa. "Analyse des propriétés stationnaires et des propriétés émergentes dans les flux d'informations changeant au cours du temps." Phd thesis, Université Henri Poincaré - Nancy I, 2009. http://tel.archives-ouvertes.fr/tel-00402644.
Full textL'apport de ce travail de thèse réside principalement dans le développement d'un modèle d'apprentissage - nommé ILoNDF - fondé sur le principe de la détection de nouveauté. L'apprentissage de ce modèle est, contrairement à sa version de départ, guidé non seulement par la nouveauté qu'apporte une donnée d'entrée mais également par la donnée elle-même. De ce fait, le modèle ILoNDF peut acquérir constamment de nouvelles connaissances relatives aux fréquences d'occurrence des données et de leurs variables, ce qui le rend moins sensible au bruit. De plus, doté d'un fonctionnement en ligne sans répétition d'apprentissage, ce modèle répond aux exigences les plus fortes liées au traitement des flux de données.
Dans un premier temps, notre travail se focalise sur l'étude du comportement du modèle ILoNDF dans le cadre général de la classification à partir d'une seule classe en partant de l'exploitation des données fortement multidimensionnelles et bruitées. Ce type d'étude nous a permis de mettre en évidence les capacités d'apprentissage pures du modèle ILoNDF vis-à-vis de l'ensemble des méthodes proposées jusqu'à présent. Dans un deuxième temps, nous nous intéressons plus particulièrement à l'adaptation fine du modèle au cadre précis du filtrage d'informations. Notre objectif est de mettre en place une stratégie de filtrage orientée-utilisateur plutôt qu'orientée-système, et ceci notamment en suivant deux types de directions. La première direction concerne la modélisation utilisateur à l'aide du modèle ILoNDF. Cette modélisation fournit une nouvelle manière de regarder le profil utilisateur en termes de critères de spécificité, d'exhaustivité et de contradiction. Ceci permet, entre autres, d'optimiser le seuil de filtrage en tenant compte de l'importance que pourrait donner l'utilisateur à la précision et au rappel. La seconde direction, complémentaire de la première, concerne le raffinement des fonctionnalités du modèle ILoNDF en le dotant d'une capacité à s'adapter à la dérive du besoin de l'utilisateur au cours du temps. Enfin, nous nous attachons à la généralisation de notre travail antérieur au cas où les données arrivant en flux peuvent être réparties en classes multiples.
Thomas, Philippe. "Contribution à l'identification de systèmes non linéaires par réseaux de neurones." Nancy 1, 1997. http://docnum.univ-lorraine.fr/public/SCD_T_1997_0030_THOMAS.pdf.
Full textThis thesis deals with the idenlificalion of dynamical non-linear ISO and MlSO systems with multilayer feedforward neural networks. Firstly, a short presentation of the non-linear identification methods is proposed and the neural network are reviewed. Secondly, the general architecture of the neural network used is more precisely defined. Some methods are presented to adapt this architecture to a particular case. These methods give the regressors and the number of neurons in the hidden layer. The relationships between neural identification and the most classical non-Iinear models are then shown. The validation criteria of non-linear models usable for the neural identification are presented. Three difficulties encountered in neural identification are investigated in the sequel. The first one is due to the initialisation of the parameters of the network. A bad choice of these initial parameters can lead to local minima very far from of the global minimum, to saturation of the hidden neurons, or to slow convergence. Two new algorithms are proposed to solve this problem and compared with others on three different examples. The slow convergence can be the result of the learning algorithm used. One algorithm is proposed to deal with this second difficulty. This algorithm is compared with the more classical RPE algorithm. This study is ends with the third studied problem which is posed by the presence of outliers in the identification data set. Lndeed, outliers can produce biases on estimated parameters. Three robust criteria are then proposed and are compared with the classical quadratic criterion on a simulation example and on a real industrial data set
Hosseini, Shahram. "Contribution à la régression non linéaire par les réseaux de neurones." Grenoble INPG, 2000. http://www.theses.fr/2000INPG0071.
Full textThiaw, Lamine. "Identification de systèmes dynamiques non linéaires par réseaux de neurones et multimodèles." Phd thesis, Université Paris XII Val de Marne, 2008. http://tel.archives-ouvertes.fr/tel-00399469.
Full textGuerry, Joris. "Reconnaissance visuelle robuste par réseaux de neurones dans des scénarios d'exploration robotique. Détecte-moi si tu peux !" Thesis, Université Paris-Saclay (ComUE), 2017. http://www.theses.fr/2017SACLX080/document.
Full textThe main objective of this thesis is visual recognition for a mobile robot in difficult conditions. We are particularly interested in neural networks which present today the best performances in computer vision. We studied the concept of method selection for the classification of 2D images by using a neural network selector to choose the best available classifier given the observed situation. This strategy works when data can be easily partitioned with respect to available classifiers, which is the case when complementary modalities are used. We have therefore used RGB-D data (2.5D) in particular applied to people detection. We propose a combination of independent neural network detectors specific to each modality (color & depth map) based on the same architecture (Faster RCNN). We share intermediate results of the detectors to allow them to complement and improve overall performance in difficult situations (luminosity loss or acquisition noise of the depth map). We are establishing new state of the art scores in the field and propose a more complex and richer data set to the community (ONERA.ROOM). Finally, we made use of the 3D information contained in the RGB-D images through a multi-view method. We have defined a strategy for generating 2D virtual views that are consistent with the 3D structure. For a semantic segmentation task, this approach artificially increases the training data for each RGB-D image and accumulates different predictions during the test. We obtain new reference results on the SUNRGBD and NYUDv2 datasets. All these works allowed us to handle in an original way 2D, 2.5D and 3D robotic data with neural networks. Whether for classification, detection and semantic segmentation, we not only validated our approaches on difficult data sets, but also brought the state of the art to a new level of performance
Artigue, Guillaume. "Prévision des crues éclair par réseaux de neurones : généralisation aux bassins non jaugés." Thesis, Montpellier 2, 2012. http://www.theses.fr/2012MON20188/document.
Full textIn the French Mediterranean regions, heavy rainfall episodes regularly occur and induce very rapid and voluminous floods called flash floods. They frequently cause fatalities and can cost more than one billion euros during only one event. In order to cope with this issue, the public authorities' implemented countermeasures in which hydrological forecasting plays an essential role.In this contexte, the French Flood Forecasting Service (called SCHAPI for Service Central d'Hydrométéorologie et d'Appui à la Prévision des Inondations) initiated the BVNE (Digital Experimental Basin, for Bassin Versant Numérique Expérimental) project in order to enhance flash flood forecasts. The present work is a part of this project and aim at three main purposes: providing flash flood forecasts on well-gauged basins, poorly gauged basins and ungauged basins.The study area chosen, the Cévennes range, concentrates the major part of these intense hydrometeorological events in France. This dissertation presents it precisely, highlighting its most hydrological-influent characteristics.With regard to the complexity of the rainfall-discharge relation in the focused basins and the difficulty experienced by the physically based models to provide precise information in forecast mode without rainfall forecasts, the use of neural networks statistical learning imposed itself in the research of operational solutions.Thus, the neural networks models were designed and applied to a basin of the Cévennes range, in the well-gauged and poorly gauged contexts. The good results obtained have been the start point of a generalization to 15 basins of the study area.For this purpose, a generalization method was developed from the model created on the gauged basin and from corrections estimated as a function of basin characteristics.The results of this method application are of good quality and open the door to numerous pats of inquiry for the future, while demonstrating again that the use of statistical learning for hydrology can be a relevant solution
Theilliol, Didier. "Identification de systèmes siso linéaires et non linéaires par réseaux de neurones multicouches." Nancy 1, 1993. http://www.theses.fr/1993NAN10261.
Full textLehalle, Charles-Albert. "Contrôle non linéaire et Réseaux de neurones formels : les Perceptrons Affines Par morceaux." Paris 6, 2005. https://tel.archives-ouvertes.fr/tel-00009592.
Full textJouini, Mohamed Soufiane. "Caractérisation des réservoirs basée sur des textures des images scanners de carottes." Thesis, Bordeaux 1, 2009. http://www.theses.fr/2009BOR13769/document.
Full textCores extracted, during wells drilling, are essential data for reservoirs characterization. A medical scanner is used for their acquisition. This feature provide high resolution images improving the capacity of interpretation. The main goal of the thesis is to establish links between these images and petrophysical data. Then parametric texture modelling can be used to achieve this goal and should provide reliable set of descriptors. A possible solution is to focus on parametric methods allowing synthesis. Even though, this method is not a proven mathematically, it provides high confidence on set of descriptors and allows interpretation into synthetic textures. In this thesis methods and algorithms were developed to achieve the following goals : 1. Segment main representative texture zones on cores. This is achieved automatically through learning and classifying textures based on parametric model. 2. Find links between scanner images and petrophysical parameters. This is achieved though calibrating and predicting petrophysical data with images (Supervised Learning Process)
Jouffroy, Guillaume. "Contrôle oscillatoire par réseau de neurones récurrents." Paris 8, 2008. http://www.theses.fr/2008PA082918.
Full textIn the control field, most of the applications need a non-oscillatory continuous control. This work focuses instead on controllers with recurrent neural networks (RNN) which generate a periodic oscillatory control. The purpose of the present work is to study stochastic optimisation methods which can be used to discover the parameters of a network so that it generates a cyclic input. First we take a look at the knowledge about biological oscillators. Tthen we describe the mathematical tools to be able to guarantee the stability oscillators. The potential of RNN, especially applied to dynamical systems being still poorly used, we propose for each method, a general detailed matrix formalization and we precise the computational complexity of the methods. We validate each method using a simple example of oscillator, and we demonstrate analytically the stability of the resulting oscillator, but also how it is robust to parameters perturbations. We then compare these different methods with these criteria and the speed of convergence. We finish this thesis with an illustration, where we take all the steps of the construction of an oscillatory neural controller, to control the axis of direction of a particular vehicle. This will let us discuss how realistic is the use of recurrent neural networks in the field of control, and propose interesting questions
Lehalle, Charles-Albert. "Le contrôle non linéaire par réseaux de neurones formels: les perceptrons affines par morceaux." Phd thesis, Université Pierre et Marie Curie - Paris VI, 2005. http://tel.archives-ouvertes.fr/tel-00009592.
Full textMazel, Johan. "Détection non supervisée d'anomalies dans les réseaux de communication." Phd thesis, INSA de Toulouse, 2011. http://tel.archives-ouvertes.fr/tel-00667654.
Full textBensetti, Mohamed. "Etude et modélisation de capteurs destinés au contrôle non destructif par courants de Foucault : mise en oeuvre de méthodes d'inversion." Paris 11, 2004. http://www.theses.fr/2004PA112291.
Full textThis thesis work approaches the problematic related ones to the non destructif testing (ndt) by eddy current, it's divided into three great parts. The first parts is consecrated to implement of the inverse model for the estimate of the physical and geometrical paramters of the tested spicemen. For this fact, we were interested in the inverse models based by neural networks. In this context two application were studied. The micro-coil are dedicated for different applications : radio frequency (rf), nuclear magnetic resonance (nmr), non destructive testing (ndt). . . Depending on the applications, these micro-coils can be used in high frequency. Acctually, the response of the micro-coil at high frequency is significanly different from their low frequency response because of the skin and proximity effects have an influence on the electrical parameters of the micro-coils. The resistance and the inductance of the winding depend on the frequency. The parasitic capacitance of the winding cannot neglected. In the second part of this work, an original method combining by 3d magnetodynamic alalysis and electrostatic analysis is presented to determine the elements of an electric equivalent circuit. The last aspect studied was devoted to developpement of an hybrid approch associating the finite element method (fem) and the boundary integral method (bim) to calculate the response of the coil in the presence of the cracks. Two types of validations were carried out in this work, a comparison of the results obtained by fem with results provided by the cea ( software civa) and a validation of the hybrid method by measurement results
Ebadzadeh, Mohamad Mehdi. "Modélisation des voies réflexes et cérébelleuses, permettant le calcul des fonctions inverses : application à la commande d'un actionneur à deux muscles pneumatiques." Paris, ENST, 2004. http://www.theses.fr/2004ENST0046.
Full textMazel, Johan. "Unsupervised network anomaly detection." Thesis, Toulouse, INSA, 2011. http://www.theses.fr/2011ISAT0024/document.
Full textAnomaly detection has become a vital component of any network in today’s Internet. Ranging from non-malicious unexpected events such as flash-crowds and failures, to network attacks such as denials-of-service and network scans, network traffic anomalies can have serious detrimental effects on the performance and integrity of the network. The continuous arising of new anomalies and attacks create a continuous challenge to cope with events that put the network integrity at risk. Moreover, the inner polymorphic nature of traffic caused, among other things, by a highly changing protocol landscape, complicates anomaly detection system's task. In fact, most network anomaly detection systems proposed so far employ knowledge-dependent techniques, using either misuse detection signature-based detection methods or anomaly detection relying on supervised-learning techniques. However, both approaches present major limitations: the former fails to detect and characterize unknown anomalies (letting the network unprotected for long periods) and the latter requires training over labeled normal traffic, which is a difficult and expensive stage that need to be updated on a regular basis to follow network traffic evolution. Such limitations impose a serious bottleneck to the previously presented problem.We introduce an unsupervised approach to detect and characterize network anomalies, without relying on signatures, statistical training, or labeled traffic, which represents a significant step towards the autonomy of networks. Unsupervised detection is accomplished by means of robust data-clustering techniques, combining Sub-Space clustering with Evidence Accumulation or Inter-Clustering Results Association, to blindly identify anomalies in traffic flows. Correlating the results of several unsupervised detections is also performed to improve detection robustness. The correlation results are further used along other anomaly characteristics to build an anomaly hierarchy in terms of dangerousness. Characterization is then achieved by building efficient filtering rules to describe a detected anomaly. The detection and characterization performances and sensitivities to parameters are evaluated over a substantial subset of the MAWI repository which contains real network traffic traces.Our work shows that unsupervised learning techniques allow anomaly detection systems to isolate anomalous traffic without any previous knowledge. We think that this contribution constitutes a great step towards autonomous network anomaly detection.This PhD thesis has been funded through the ECODE project by the European Commission under the Framework Programme 7. The goal of this project is to develop, implement, and validate experimentally a cognitive routing system that meet the challenges experienced by the Internet in terms of manageability and security, availability and accountability, as well as routing system scalability and quality. The concerned use case inside the ECODE project is network anomaly
Forgez, Christophe. "Méthodologie de modélisation et de commande par réseaux de neurones pour des dispositifs électrotechniques non linéaires." Lille 1, 1998. https://pepite-depot.univ-lille.fr/LIBRE/Th_Num/1998/50376-1998-309.pdf.
Full textTertois, Sylvain. "Réduction des effets des non-linéarités dans une modulation multiporteuse à l'aire de réseaux de neurones." Phd thesis, Université Rennes 1, 2003. http://tel.archives-ouvertes.fr/tel-00004015.
Full textTout d'abord, le mémoire commence par une introduction aux communications numériques et en particulier à la modulation OFDM. Aujourd'hui, plusieurs standards reposent sur cette technique de transmission, en particulier en raison de la simplicité de l'égalisation du canal, et donc la possibilité de transmettre avec plus d'efficacité des données sur des canaux multitrajets. Cependant le signal OFDM temporel est particulièrement sensible aux non-linéarités présentes dans l'amplificateur d'émission et diverses techniques sont étudiées pour diminuer ces effets.
Ensuite, les réseaux de neurones sont présentés, ainsi que leur utilisation dans le domaine de l'approximation de fonctions. Après avoir décrit les deux modèles de réseaux de neurones les plus courants, les réseaux d'ordre supérieur, tels que le RPN, sont introduits. Les techniques d'apprentissage de ces différentes architectures de réseaux de neurones sont également décrites.
Dans les différents correcteurs étudiés dans cette thèse, le réseau de neurones est placé dans le récepteur, après l'égalisation de canal. Son objectif est de corriger le signal reçu afin de compenser les effets des non-linéarités. Dans un premier temps le réseau de neurones est placé dans le domaine fréquentiel. Dans un système OFDM à 4 porteuses avec une modulation MAQ16, un amplificateur de type SSPA, un recul de 0 dB et pour un taux d'erreur binaire de 10-2, le correcteur avec un réseau RPN apporte un gain de 1,5 dB de rapport signal sur bruit. Cependant des difficultés apparaissent durant la phase d'apprentissage du réseau de neurones avec un nombre de porteuses supérieur.
Pour palier ce défaut, les réseaux de neurones décrits précédemment sont simplifiés en étant placés dans le domaine temporel. Ce système est plus proche des solutions déjà proposées pour la compensation des non-linéarités dans une modulation monoporteuse, avec toutefois des différences au niveau de l'égalisation du canal et de la nature de la fonction que doit accomplir le réseau de neurones. Un correcteur basé sur un réseau RPN a montré de très bonnes performances, même en augmentant le nombre de porteuses. Un gain de 8 dB a été mesuré pour un taux d'erreur binaire de 10-2 dans un système OFDM à 48 porteuses, une modulation MAQ16 et un amplificateur de type SSPA avec un recul de 0 dB. Le système présenté permet donc dans ces conditions de diviser la puissance de l'amplificateur, et donc sa consommation d'énergie, par un facteur supérieur à 4 tout en conservant la même qualité de transmission.
Le correcteur à RPN dans le domaine temporel est ensuite simulé sur un canal multitrajet, afin de vérifier que la compensation reste efficace dans le cas d'un canal sévère. Enfin les deux approches proposées (fréquentielle et temporelle) sont comparées, au niveau des performances obtenues et de la puissance de calcul nécessaire dans le récepteur. Une comparaison avec une autre approche proposée dans la littérature est également présentée. Le correcteur temporel basé sur un RPN est bien moins complexe que le système cité, au détriment d'une légère dégradation des performances.
Ce mémoire se conclut par quelques perspectives de recherche pouvant prolonger les travaux accomplis durant cette thèse.
Zhang, Jiantong. "Estimation non paramétrique par réseaux de neurones pour des processus et estimation de paramètres de position." Paris 6, 2000. http://www.theses.fr/2000PA066485.
Full textDjouani, Karim. "Contribution à la commande dynamique des navires. Commande optimale non-linéaire et réseaux de neurones artificiels." Paris 12, 1994. http://www.theses.fr/1994PA120043.
Full textTertois, Sylvain. "Réduction des effets de non-linéarités dans une modulation multiporteuse à l'aide de réseaux de neurones." Rennes 1, 2003. https://tel.archives-ouvertes.fr/tel-00004015.
Full textHarkouss, Youssef. "Application de réseaux de neurones à la modélisation de composants et de dispositifs microondes non linéaires." Limoges, 1998. http://www.theses.fr/1998LIMO0040.
Full textGraux, François. "Méthodologie de modélisation boîte noire de circuits hyperfréquences non linéaires par réseaux de neurones : applications au radar." Lille 1, 2001. https://pepite-depot.univ-lille.fr/RESTREINT/Th_Num/2001/50376-2001-47.pdf.
Full textJemeï, Samir. "Modélisation d'une Pile à Combustible de type PEM par Réseaux de Neurones." Phd thesis, Université de Franche-Comté, 2004. http://tel.archives-ouvertes.fr/tel-00777611.
Full textMba, Megner Bruno. "Étude de l'application des réseaux de neurones à la reconstitution des signaux pour un système dynamique non linéaire." Thèse, Université du Québec à Trois-Rivières, 1998. http://depot-e.uqtr.ca/3561/1/000654592.pdf.
Full textAgnaou, Youssef Joseph. "Analyse statistique de données de croissance humaine : estimation et ajustement paramétriques, non paramétriques, et par réseaux de neurones." Bordeaux 1, 2001. http://www.theses.fr/2001BOR12404.
Full textCazé, Romain. "Le rôle des sommations non-linéaires dendritiques dans la puissance computationnelle du neurone." Paris 7, 2012. http://www.theses.fr/2012PA077238.
Full textSeminal computational models of the neuron assume that excitatory post-synaptic inputs (EPSPs) sum linearly in dendrites. Nevertheless, the sum of multiple EPSPs can be larger than their arithmetic sum, a superlinear summation. The so-called dendritic spike. An impact of dendritic spikes on computation remains a malter of debate. Moreover, the sum of multiple of EPSPs can also be smaller than their arithmetic sum these saturations are sometime presented as a glitch which should be corrected by dendritic spikes. L provide here arguments against these daims, I show that dendritic saturations as well as dendritic spikes, even when they cannot directly make the neuron fire, enhance single neuron computation. I use a binary neuron models to demonstrate that a single dendritic non-linearity, either spiking or saturating, combined with somatic non-linearity, enables a neuron linearly non-separable functions. I coin these functions as spatial functions because the neuron's output depends more on the labeling of the active inputs than on their number. Secondly, we show that realistic biophysical models of the neuron are capable of computing spatial functions. Within these models the dendritic and somatic non-linearity are tightly coupled. We use this biophysical model to predict that some neurons will be more likely to fire when inputs are scattered over their dendritic tree than when they are clustered, we predict that such a neuron is capable of computing spatial functions. These results suggest a new perspective on neural networks, suggesting for instance that memory can be stored in the neurons themselves
Benoît-Marand, François. "Modélisation et identification des systèmes non linéaires par réseaux de neurones à temps continu : application à la modélisation des interfaces de diffusion non linéaires." Poitiers, 2007. http://www.theses.fr/2007POIT2274.
Full textThis thesis presents a new model for the identification of nonlinear systems : continuous time neural networks (RNTC). These structures employ networks of formal neurons to approach the nonlinear laws that control the system but, contrary to the neural networks models presented in the literature, our model deals the problem in continuous time. Whatever, through various applications, we show that the model allows us to identify various nonlinear processes with a high accuracy. Moreover, in using a model reduction stage, it is possible to revert, from the neural network model, to the characteristic values of the system. Finally, we indicate how to adapt the continuous time neural network model to the case of fractionnal systems and we consider the problem of identification of diffusive nonlinear interfaces. By introducing a new operator of fractional integration, and by integrating it into the continuous time neural network model, we show how to approach the temporal behavior of these particular systems
Abdulkader, Hasan. "Applications de réseaux de neurones à des chaînes de transmission numérique par satellite-gradient naturel." Toulouse, INPT, 2003. http://www.theses.fr/2003INPT009H.
Full text