Stimulus statistics shape oscillations in nonlinear recurrent neural networks.

Rhythmic activity plays a central role in neural computations and brain functions ranging from homeostasis to attention, as well as in neurological and neuropsychiatric disorders. Despite this pervasiveness, little is known about the mechanisms whereby the frequency and power of oscillatory activity are modulated, and how they reflect the inputs received by neurons. Numerous studies have reported input-dependent fluctuations in peak frequency and power (as well as couplings across these features). However, it remains unresolved what mediates these spectral shifts among neural populations. Extending previous findings regarding stochastic nonlinear systems and experimental observations, we provide analytical insights regarding oscillatory responses of neural populations to stimulation from either endogenous or exogenous origins. Using a deceptively simple yet sparse and randomly connected network of neurons, we show how spiking inputs can reliably modulate the peak frequency and power expressed by synchronous neural populations without any changes in circuitry. Our results reveal that a generic, non-nonlinear and input-induced mechanism can robustly mediate these spectral fluctuations, and thus provide a framework in which inputs to the neurons bidirectionally regulate both the frequency and power expressed by synchronous populations. Theoretical and computational analysis of the ensuing spectral fluctuations was found to reflect the underlying dynamics of the input stimuli driving the neurons. Our results provide insights regarding a generic mechanism supporting spectral transitions observed across cortical networks and spanning multiple frequency bands.

Extrasynaptic neurotransmission in the modulation of brain function. Focus on the striatal neuronal glial networks

Extrasynaptic neurotransmission is an important short distance form of volume transmission (VT) and describes the extracellular diffusion of transmitters and modulators after synaptic spillover or extrasynaptic release in the local circuit regions binding to and activating mainly extrasynaptic neuronal and glial receptors in the neuroglial networks of the brain. Receptor-receptor interactions in G protein-coupled receptor (GPCR) heteromers play a major role, on dendritic spines and nerve terminals including glutamate synapses, in the integrative processes of the extrasynaptic signaling. Heteromeric complexes between GPCR and ion-channel receptors play a special role in the integration of the synaptic and extrasynaptic signals. Changes in extracellular concentrations of the classical synaptic neurotransmitters glutamate and GABA found with microdialysis is likely an expression of the activity of the neuron-astrocyte unit of the brain and can be used as an index of VT-mediated actions of these two neurotransmitters in the brain. Thus, the activity of neurons may be functionally linked to the activity of astrocytes, which may release glutamate and GABA to the extracellular space where extrasynaptic glutamate and GABA receptors do exist. Wiring transmission (WT) and VT are fundamental properties of all neurons of the CNS but the balance between WT and VT varies from one nerve cell population to the other. The focus is on the striatal cellular networks, and the WT and VT and their integration via receptor heteromers are described in the GABA projection neurons, the glutamate, dopamine, 5-hydroxytryptamine (5-HT) and histamine striatal afferents, the cholinergic interneurons, and different types of GABA interneurons. In addition, the role in these networks of VT signaling of the energy-dependent modulator adenosine and of endocannabinoids mainly formed in the striatal projection neurons will be underlined to understand the communication in the striatal cellular networks

Spatial prediction of monthly wind speeds in complex terrain with adaptive general regression neural networks

This paper presents the general regression neural networks (GRNN) as a nonlinear regression method for the interpolation of monthly wind speeds in complex Alpine orography. GRNN is trained using data coming from Swiss meteorological networks to learn the statistical relationship between topographic features and wind speed. The terrain convexity, slope and exposure are considered by extracting features from the digital elevation model at different spatial scales using specialised convolution filters. A database of gridded monthly wind speeds is then constructed by applying GRNN in prediction mode during the period 1968-2008. This study demonstrates that using topographic features as inputs in GRNN significantly reduces cross-validation errors with respect to low-dimensional models integrating only geographical coordinates and terrain height for the interpolation of wind speed. The spatial predictability of wind speed is found to be lower in summer than in winter due to more complex and weaker wind-topography relationships. The relevance of these relationships is studied using an adaptive version of the GRNN algorithm which allows to select the useful terrain features by eliminating the noisy ones. This research provides a framework for extending the low-dimensional interpolation models to high-dimensional spaces by integrating additional features accounting for the topographic conditions at multiple spatial scales. Copyright (c) 2012 Royal Meteorological Society.

Integrating defeasible argumentation with fuzzy ART neural networks for pattern classification

Many classification systems rely on clustering techniques in which a collection of training examples is provided as an input, and a number of clusters c1,...cm modelling some concept C results as an output, such that every cluster ci is labelled as positive or negative. Given a new, unlabelled instance enew, the above classification is used to determine to which particular cluster ci this new instance belongs. In such a setting clusters can overlap, and a new unlabelled instance can be assigned to more than one cluster with conflicting labels. In the literature, such a case is usually solved non-deterministically by making a random choice. This paper presents a novel, hybrid approach to solve this situation by combining a neural network for classification along with a defeasible argumentation framework which models preference criteria for performing clustering.

Modelling bird species richness with neural networks for forest landscape management in NE Spain

Para preservar la biodiversidad de los ecosistemas forestales de la Europa mediterránea en escenarios actuales y futuros de cambio global mediante una gestión forestal sostenible es necesario determinar cómo influye el medio ambiente y las propias características de los bosques sobre la biodiversidad que éstos albergan. Con este propósito, se analizó la influencia de diferentes factores ambientales y de estructura y composición del bosque sobre la riqueza de aves forestales a escala 1 × 1 km en Cataluña (NE de España). Se construyeron modelos univariantes y multivariantes de redes neuronales para respectivamente explorar la respuesta individual a las variables y obtener un modelo parsimonioso (ecológicamente interpretable) y preciso. La superficie de bosque (con una fracción de cabida cubierta superior a 5%), la fracción de cabida cubierta media, la temperatura anual y la precipitación estival medias fueron los mejores predictores de la riqueza de aves forestales. La red neuronal multivariante obtenida tuvo una buena capacidad de generalización salvo en las localidades con una mayor riqueza. Además, los bosques con diferentes grados de apertura del dosel arbóreo, más maduros y más diversos en cuanto a su composición de especies arbóreas se asociaron de forma positiva con una mayor riqueza de aves forestales. Finalmente, se proporcionan directrices de gestión para la planificación forestal que permitan promover la diversidad ornítica en esta región de la Europa mediterránea.

Generalization transitions in Hidden-Layer neural networks for third-order feature discrimination

Stochastic learning processes for a specific feature detector are studied. This technique is applied to nonsmooth multilayer neural networks requested to perform a discrimination task of order 3 based on the ssT-block¿ssC-block problem. Our system proves to be capable of achieving perfect generalization, after presenting finite numbers of examples, by undergoing a phase transition. The corresponding annealed theory, which involves the Ising model under external field, shows good agreement with Monte Carlo simulations.

Application of artificial neural networks as an alternative to volumetric water balance in drip irrigation management in watermelon crop

Precision irrigation seeks to establish strategies which achieve an efficient ratio between the volume of water used (reduction in input) and the productivity obtained (increase in production). There are several studies in the literature on strategies for achieving this efficiency, such as those dealing with the method of volumetric water balance (VWB). However, it is also of great practical and economic interest to set up versatile implementations of irrigation strategies that: (i) maintain the performance obtained with other implementations, (ii) rely on few computational resources, (iii) adapt well to field conditions, and (iv) allow easy modification of the irrigation strategy. In this study, such characteristics are achieved when using an Artificial Neural Network (ANN) to determine the period of irrigation for a watermelon crop in the Irrigation Perimeter of the Lower Acaraú, in the state of Ceará, Brazil. The Volumetric Water Balance was taken as the standard for comparing the management carried out with the proposed implementation of ANN. The statistical analysis demonstrates the effectiveness of the proposed management, which is able to replace VWB as a strategy in automation.

Training executive functions in the laboratory and in real life : cognitive consequences of computer-based exercises and bilingualism

The question of the trainability of executive functions and the impact of such training on related cognitive skills has stirred considerable research interest. Despite a number of studies investigating this, the question has not yet been solved. The general aim of this thesis was to investigate two very different types of training of executive functions: laboratory-based computerized training (Studies I-III) and realworld training through bilingualism (Studies IV-V). Bilingualism as a kind of training of executive functions is based on the idea that managing two languages requires executive resources, and previous studies have suggested a bilingual advantage in executive functions. Three executive functions were studied in the present thesis: updating of working memory (WM) contents, inhibition of irrelevant information, and shifting between tasks and mental sets. Studies I-III investigated the effects of computer-based training of WM updating (Study I), inhibition (Study II), and set shifting (Study III) in healthy young adults. All studies showed increased performance on the trained task. More importantly, improvement on an untrained task tapping the trained executive function (near transfer) was seen in Study I and II. None of the three studies showed improvement on untrained tasks tapping some other cognitive function (far transfer) as a result of training. Study I also used PET to investigate the effects of WM updating training on a neurotransmitter closely linked to WM, namely dopamine. The PET results revealed increased striatal dopamine release during WM updating performance as a result of training. Study IV investigated the ability to inhibit task-irrelevant stimuli in bilinguals and monolinguals by using a dichotic listening task. The results showed that the bilinguals exceeded the monolinguals in inhibiting task-irrelevant information. Study V introduced a new, complementary research approach to study the bilingual executive advantage and its underlying mechanisms. To circumvent the methodological problems related to natural groups design, this approach focuses only on bilinguals and examines whether individual differences in bilingual behavior correlate with executive task performances. Using measures that tap the three above-entioned executive functions, the results suggested that more frequent language switching was associated with better set shifting skills, and earlier acquisition of the second language was related to better inhibition skills. In conclusion, the present behavioral results showed that computer-based training of executive functions can improve performance on the trained task and on closely related tasks, but does not yield a more general improvement of cognitive skills. Moreover, the functional neuroimaging results reveal that WM training modulates striatal dopaminergic function, speaking for training-induced neural plasticity in this important neurotransmitter system. With regard to bilingualism, the results provide further support to the idea that bilingualism can enhance executive functions. In addition, the new complementary research approach proposed here provides some clues as to which aspects of everyday bilingual behavior may be related to the advantage in executive functions in bilingual individuals.

Classification of antimicrobial resistance using artificial neural networks and the relationship of 38 genes associated with the virulence of Escherichia coli isolates from broilers

Avian pathogenic Escherichia coli (APEC) is responsible for various pathological processes in birds and is considered as one of the principal causes of morbidity and mortality, associated with economic losses to the poultry industry. The objective of this study was to demonstrate that it is possible to predict antimicrobial resistance of 256 samples (APEC) using 38 different genes responsible for virulence factors, through a computer program of artificial neural networks (ANNs). A second target was to find the relationship between (PI) pathogenicity index and resistance to 14 antibiotics by statistical analysis. The results showed that the RNAs were able to make the correct classification of the behavior of APEC samples with a range from 74.22 to 98.44%, and make it possible to predict antimicrobial resistance. The statistical analysis to assess the relationship between the pathogenic index (PI) and resistance against 14 antibiotics showed that these variables are independent, i.e. peaks in PI can happen without changing the antimicrobial resistance, or the opposite, changing the antimicrobial resistance without a change in PI.

Synergistic control of forearm based on accelerometer data and artificial neural networks

In the present study, we modeled a reaching task as a two-link mechanism. The upper arm and forearm motion trajectories during vertical arm movements were estimated from the measured angular accelerations with dual-axis accelerometers. A data set of reaching synergies from able-bodied individuals was used to train a radial basis function artificial neural network with upper arm/forearm tangential angular accelerations. The trained radial basis function artificial neural network for the specific movements predicted forearm motion from new upper arm trajectories with high correlation (mean, 0.9149-0.941). For all other movements, prediction was low (range, 0.0316-0.8302). Results suggest that the proposed algorithm is successful in generalization over similar motions and subjects. Such networks may be used as a high-level controller that could predict forearm kinematics from voluntary movements of the upper arm. This methodology is suitable for restoring the upper limb functions of individuals with motor disabilities of the forearm, but not of the upper arm. The developed control paradigm is applicable to upper-limb orthotic systems employing functional electrical stimulation. The proposed approach is of great significance particularly for humans with spinal cord injuries in a free-living environment. The implication of a measurement system with dual-axis accelerometers, developed for this study, is further seen in the evaluation of movement during the course of rehabilitation. For this purpose, training-related changes in synergies apparent from movement kinematics during rehabilitation would characterize the extent and the course of recovery. As such, a simple system using this methodology is of particular importance for stroke patients. The results underlie the important issue of upper-limb coordination.

Artificial neural networks (ANN): prediction of sensory measurements from instrumental data

The objective of this study was to predict by means of Artificial Neural Network (ANN), multilayer perceptrons, the texture attributes of light cheesecurds perceived by trained judges based on instrumental texture measurements. Inputs to the network were the instrumental texture measurements of light cheesecurd (imitative and fundamental parameters). Output variables were the sensory attributes consistency and spreadability. Nine light cheesecurd formulations composed of different combinations of fat and water were evaluated. The measurements obtained by the instrumental and sensory analyses of these formulations constituted the data set used for training and validation of the network. Network training was performed using a back-propagation algorithm. The network architecture selected was composed of 8-3-9-2 neurons in its layers, which quickly and accurately predicted the sensory texture attributes studied, showing a high correlation between the predicted and experimental values for the validation data set and excellent generalization ability, with a validation RMSE of 0.0506.

Pretraining Convolutional Neural Networks for Visual Recognition

Convolutional Neural Networks (CNN) have become the state-of-the-art methods on many large scale visual recognition tasks. For a lot of practical applications, CNN architectures have a restrictive requirement: A huge amount of labeled data are needed for training. The idea of generative pretraining is to obtain initial weights of the network by training the network in a completely unsupervised way and then fine-tune the weights for the task at hand using supervised learning. In this thesis, a general introduction to Deep Neural Networks and algorithms are given and these methods are applied to classification tasks of handwritten digits and natural images for developing unsupervised feature learning. The goal of this thesis is to find out if the effect of pretraining is damped by recent practical advances in optimization and regularization of CNN. The experimental results show that pretraining is still a substantial regularizer, however, not a necessary step in training Convolutional Neural Networks with rectified activations. On handwritten digits, the proposed pretraining model achieved a classification accuracy comparable to the state-of-the-art methods.

Incorporating complex cells into neural networks for pattern classification

Dans le domaine des neurosciences computationnelles, l'hypothèse a été émise que le système visuel, depuis la rétine et jusqu'au cortex visuel primaire au moins, ajuste continuellement un modèle probabiliste avec des variables latentes, à son flux de perceptions. Ni le modèle exact, ni la méthode exacte utilisée pour l'ajustement ne sont connus, mais les algorithmes existants qui permettent l'ajustement de tels modèles ont besoin de faire une estimation conditionnelle des variables latentes. Cela nous peut nous aider à comprendre pourquoi le système visuel pourrait ajuster un tel modèle; si le modèle est approprié, ces estimé conditionnels peuvent aussi former une excellente représentation, qui permettent d'analyser le contenu sémantique des images perçues. Le travail présenté ici utilise la performance en classification d'images (discrimination entre des types d'objets communs) comme base pour comparer des modèles du système visuel, et des algorithmes pour ajuster ces modèles (vus comme des densités de probabilité) à des images. Cette thèse (a) montre que des modèles basés sur les cellules complexes de l'aire visuelle V1 généralisent mieux à partir d'exemples d'entraînement étiquetés que les réseaux de neurones conventionnels, dont les unités cachées sont plus semblables aux cellules simples de V1; (b) présente une nouvelle interprétation des modèles du système visuels basés sur des cellules complexes, comme distributions de probabilités, ainsi que de nouveaux algorithmes pour les ajuster à des données; et (c) montre que ces modèles forment des représentations qui sont meilleures pour la classification d'images, après avoir été entraînés comme des modèles de probabilités. Deux innovations techniques additionnelles, qui ont rendu ce travail possible, sont également décrites : un algorithme de recherche aléatoire pour sélectionner des hyper-paramètres, et un compilateur pour des expressions mathématiques matricielles, qui peut optimiser ces expressions pour processeur central (CPU) et graphique (GPU).

Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

On Recurrent and Deep Neural Networks

L'apprentissage profond est un domaine de recherche en forte croissance en apprentissage automatique qui est parvenu à des résultats impressionnants dans différentes tâches allant de la classification d'images à la parole, en passant par la modélisation du langage. Les réseaux de neurones récurrents, une sous-classe d'architecture profonde, s'avèrent particulièrement prometteurs. Les réseaux récurrents peuvent capter la structure temporelle dans les données. Ils ont potentiellement la capacité d'apprendre des corrélations entre des événements éloignés dans le temps et d'emmagasiner indéfiniment des informations dans leur mémoire interne. Dans ce travail, nous tentons d'abord de comprendre pourquoi la profondeur est utile. Similairement à d'autres travaux de la littérature, nos résultats démontrent que les modèles profonds peuvent être plus efficaces pour représenter certaines familles de fonctions comparativement aux modèles peu profonds. Contrairement à ces travaux, nous effectuons notre analyse théorique sur des réseaux profonds acycliques munis de fonctions d'activation linéaires par parties, puisque ce type de modèle est actuellement l'état de l'art dans différentes tâches de classification. La deuxième partie de cette thèse porte sur le processus d'apprentissage. Nous analysons quelques techniques d'optimisation proposées récemment, telles l'optimisation Hessian free, la descente de gradient naturel et la descente des sous-espaces de Krylov. Nous proposons le cadre théorique des méthodes à région de confiance généralisées et nous montrons que plusieurs de ces algorithmes développés récemment peuvent être vus dans cette perspective. Nous argumentons que certains membres de cette famille d'approches peuvent être mieux adaptés que d'autres à l'optimisation non convexe. La dernière partie de ce document se concentre sur les réseaux de neurones récurrents. Nous étudions d'abord le concept de mémoire et tentons de répondre aux questions suivantes: Les réseaux récurrents peuvent-ils démontrer une mémoire sans limite? Ce comportement peut-il être appris? Nous montrons que cela est possible si des indices sont fournis durant l'apprentissage. Ensuite, nous explorons deux problèmes spécifiques à l'entraînement des réseaux récurrents, à savoir la dissipation et l'explosion du gradient. Notre analyse se termine par une solution au problème d'explosion du gradient qui implique de borner la norme du gradient. Nous proposons également un terme de régularisation conçu spécifiquement pour réduire le problème de dissipation du gradient. Sur un ensemble de données synthétique, nous montrons empiriquement que ces mécanismes peuvent permettre aux réseaux récurrents d'apprendre de façon autonome à mémoriser des informations pour une période de temps indéfinie. Finalement, nous explorons la notion de profondeur dans les réseaux de neurones récurrents. Comparativement aux réseaux acycliques, la définition de profondeur dans les réseaux récurrents est souvent ambiguë. Nous proposons différentes façons d'ajouter de la profondeur dans les réseaux récurrents et nous évaluons empiriquement ces propositions.