Population-based continuous optimization, probabilistic modelling and mean shift

Evolutionary algorithms perform optimization using a population of sample solution points. An interesting development has been to view population-based optimization as the process of evolving an explicit, probabilistic model of the search space. This paper investigates a formal basis for continuous, population-based optimization in terms of a stochastic gradient descent on the Kullback-Leibler divergence between the model probability density and the objective function, represented as an unknown density of assumed form. This leads to an update rule that is related and compared with previous theoretical work, a continuous version of the population-based incremental learning algorithm, and the generalized mean shift clustering framework. Experimental results are presented that demonstrate the dynamics of the new algorithm on a set of simple test problems.

On-line learning in multilayer neural networks

We present an analytic solution to the problem of on-line gradient-descent learning for two-layer neural networks with an arbitrary number of hidden units in both teacher and student networks. The technique, demonstrated here for the case of adaptive input-to-hidden weights, becomes exact as the dimensionality of the input space increases.

Globally optimal parameters for on-line learning in multilayer neural networks

We present a framework for calculating globally optimal parameters, within a given time frame, for on-line learning in multilayer neural networks. We demonstrate the capability of this method by computing optimal learning rates in typical learning scenarios. A similar treatment allows one to determine the relevance of related training algorithms based on modifications to the basic gradient descent rule as well as to compare different training methods.

On-line learning with adaptive back-propagation in two-layer networks

An adaptive back-propagation algorithm parameterized by an inverse temperature 1/T is studied and compared with gradient descent (standard back-propagation) for on-line learning in two-layer neural networks with an arbitrary number of hidden units. Within a statistical mechanics framework, we analyse these learning algorithms in both the symmetric and the convergence phase for finite learning rates in the case of uncorrelated teachers of similar but arbitrary length T. These analyses show that adaptive back-propagation results generally in faster training by breaking the symmetry between hidden units more efficiently and by providing faster convergence to optimal generalization than gradient descent.

Globally optimal learning rates in multilayer neural networks

A method for calculating the globally optimal learning rate in on-line gradient-descent training of multilayer neural networks is presented. The method is based on a variational approach which maximizes the decrease in generalization error over a given time frame. We demonstrate the method by computing optimal learning rates in typical learning scenarios. The method can also be employed when different learning rates are allowed for different parameter vectors as well as to determine the relevance of related training algorithms based on modifications to the basic gradient descent rule.

Incorporating curvature information into on-line learning

We analyse the dynamics of a number of second order on-line learning algorithms training multi-layer neural networks, using the methods of statistical mechanics. We first consider on-line Newton's method, which is known to provide optimal asymptotic performance. We determine the asymptotic generalization error decay for a soft committee machine, which is shown to compare favourably with the result for standard gradient descent. Matrix momentum provides a practical approximation to this method by allowing an efficient inversion of the Hessian. We consider an idealized matrix momentum algorithm which requires access to the Hessian and find close correspondence with the dynamics of on-line Newton's method. In practice, the Hessian will not be known on-line and we therefore consider matrix momentum using a single example approximation to the Hessian. In this case good asymptotic performance may still be achieved, but the algorithm is now sensitive to parameter choice because of noise in the Hessian estimate. On-line Newton's method is not appropriate during the transient learning phase, since a suboptimal unstable fixed point of the gradient descent dynamics becomes stable for this algorithm. A principled alternative is to use Amari's natural gradient learning algorithm and we show how this method provides a significant reduction in learning time when compared to gradient descent, while retaining the asymptotic performance of on-line Newton's method.

Transients and asymptotics of natural gradient learning

We analyse natural gradient learning in a two-layer feed-forward neural network using a statistical mechanics framework which is appropriate for large input dimension. We find significant improvement over standard gradient descent in both the transient and asymptotic phases of learning.

Natural gradient matrix momentum

Natural gradient learning is an efficient and principled method for improving on-line learning. In practical applications there will be an increased cost required in estimating and inverting the Fisher information matrix. We propose to use the matrix momentum algorithm in order to carry out efficient inversion and study the efficacy of a single step estimation of the Fisher information matrix. We analyse the proposed algorithm in a two-layer network, using a statistical mechanics framework which allows us to describe analytically the learning dynamics, and compare performance with true natural gradient learning and standard gradient descent.

A Generalized Fuzzy Linguistic Model for Predicting Component Concentrations in an Optical Gas Sensing System

Motivated by environmental protection concerns, monitoring the flue gas of thermal power plant is now often mandatory due to the need to ensure that emission levels stay within safe limits. Optical based gas sensing systems are increasingly employed for this purpose, with regression techniques used to relate gas optical absorption spectra to the concentrations of specific gas components of interest (NOx, SO2 etc.). Accurately predicting gas concentrations from absorption spectra remains a challenging problem due to the presence of nonlinearities in the relationships and the high-dimensional and correlated nature of the spectral data. This article proposes a generalized fuzzy linguistic model (GFLM) to address this challenge. The GFLM is made up of a series of “If-Then” fuzzy rules. The absorption spectra are input variables in the rule antecedent. The rule consequent is a general nonlinear polynomial function of the absorption spectra. Model parameters are estimated using least squares and gradient descent optimization algorithms. The performance of GFLM is compared with other traditional prediction models, such as partial least squares, support vector machines, multilayer perceptron neural networks and radial basis function networks, for two real flue gas spectral datasets: one from a coal-fired power plant and one from a gas-fired power plant. The experimental results show that the generalized fuzzy linguistic model has good predictive ability, and is competitive with alternative approaches, while having the added advantage of providing an interpretable model.

A generalized fuzzy linguistic model for predicting component concentrations in an optical gas sensing system

Motivated by environmental protection concerns, monitoring the flue gas of thermal power plant is now often mandatory due to the need to ensure that emission levels stay within safe limits. Optical based gas sensing systems are increasingly employed for this purpose, with regression techniques used to relate gas optical absorption spectra to the concentrations of specific gas components of interest (NO_x, SO₂ etc.). Accurately predicting gas concentrations from absorption spectra remains a challenging problem due to the presence of nonlinearities in the relationships and the high-dimensional and correlated nature of the spectral data. This article proposes a generalized fuzzy linguistic model (GFLM) to address this challenge. The GFLM is made up of a series of “If-Then” fuzzy rules. The absorption spectra are input variables in the rule antecedent. The rule consequent is a general nonlinear polynomial function of the absorption spectra. Model parameters are estimated using least squares and gradient descent optimization algorithms. The performance of GFLM is compared with other traditional prediction models, such as partial least squares, support vector machines, multilayer perceptron neural networks and radial basis function networks, for two real flue gas spectral datasets: one from a coal-fired power plant and one from a gas-fired power plant. The experimental results show that the generalized fuzzy linguistic model has good predictive ability, and is competitive with alternative approaches, while having the added advantage of providing an interpretable model.

Spatio-temporal coverage optimization of sensor networks

Les réseaux de capteurs sont formés d’un ensemble de dispositifs capables de prendre individuellement des mesures d’un environnement particulier et d’échanger de l’information afin d’obtenir une représentation de haut niveau sur les activités en cours dans la zone d’intérêt. Une telle détection distribuée, avec de nombreux appareils situés à proximité des phénomènes d’intérêt, est pertinente dans des domaines tels que la surveillance, l’agriculture, l’observation environnementale, la surveillance industrielle, etc. Nous proposons dans cette thèse plusieurs approches pour effectuer l’optimisation des opérations spatio-temporelles de ces dispositifs, en déterminant où les placer dans l’environnement et comment les contrôler au fil du temps afin de détecter les cibles mobiles d’intérêt. La première nouveauté consiste en un modèle de détection réaliste représentant la couverture d’un réseau de capteurs dans son environnement. Nous proposons pour cela un modèle 3D probabiliste de la capacité de détection d’un capteur sur ses abords. Ce modèle inègre également de l’information sur l’environnement grâce à l’évaluation de la visibilité selon le champ de vision. À partir de ce modèle de détection, l’optimisation spatiale est effectuée par la recherche du meilleur emplacement et l’orientation de chaque capteur du réseau. Pour ce faire, nous proposons un nouvel algorithme basé sur la descente du gradient qui a été favorablement comparée avec d’autres méthodes génériques d’optimisation «boites noires» sous l’aspect de la couverture du terrain, tout en étant plus efficace en terme de calculs. Une fois que les capteurs placés dans l’environnement, l’optimisation temporelle consiste à bien couvrir un groupe de cibles mobiles dans l’environnement. D’abord, on effectue la prédiction de la position future des cibles mobiles détectées par les capteurs. La prédiction se fait soit à l’aide de l’historique des autres cibles qui ont traversé le même environnement (prédiction à long terme), ou seulement en utilisant les déplacements précédents de la même cible (prédiction à court terme). Nous proposons de nouveaux algorithmes dans chaque catégorie qui performent mieux ou produits des résultats comparables par rapport aux méthodes existantes. Une fois que les futurs emplacements de cibles sont prédits, les paramètres des capteurs sont optimisés afin que les cibles soient correctement couvertes pendant un certain temps, selon les prédictions. À cet effet, nous proposons une méthode heuristique pour faire un contrôle de capteurs, qui se base sur les prévisions probabilistes de trajectoire des cibles et également sur la couverture probabiliste des capteurs des cibles. Et pour terminer, les méthodes d’optimisation spatiales et temporelles proposées ont été intégrées et appliquées avec succès, ce qui démontre une approche complète et efficace pour l’optimisation spatio-temporelle des réseaux de capteurs.

A deep learning model via thermodynamics and condensed matter physics

Deep Learning architectures give brilliant results in a large variety of fields, but a comprehensive theoretical description of their inner functioning is still lacking. In this work, we try to understand the behavior of neural networks by modelling in the frameworks of Thermodynamics and Condensed Matter Physics. We approach neural networks as in a real laboratory and we measure the frequency spectrum and the entropy of the weights of the trained model. The stochasticity of the training occupies a central role in the dynamics of the weights and makes it difficult to assimilate neural networks to simple physical systems. However, the analogy with Thermodynamics and the introduction of a well defined temperature leads us to an interesting result: if we eliminate from a CNN the "hottest" filters, the performance of the model remains the same, whereas, if we eliminate the "coldest" ones, the performance gets drastically worst. This result could be exploited in the realization of a training loop which eliminates the filters that do not contribute to loss reduction. In this way, the computational cost of the training will be lightened and more importantly this would be done by following a physical model. In any case, beside important practical applications, our analysis proves that a new and improved modeling of Deep Learning systems can pave the way to new and more efficient algorithms.

Improving the tracking capability of adaptive filters via convex combination

As is well known, Hessian-based adaptive filters (such as the recursive-least squares algorithm (RLS) for supervised adaptive filtering, or the Shalvi-Weinstein algorithm (SWA) for blind equalization) converge much faster than gradient-based algorithms [such as the least-mean-squares algorithm (LMS) or the constant-modulus algorithm (CMA)]. However, when the problem is tracking a time-variant filter, the issue is not so clear-cut: there are environments for which each family presents better performance. Given this, we propose the use of a convex combination of algorithms of different families to obtain an algorithm with superior tracking capability. We show the potential of this combination and provide a unified theoretical model for the steady-state excess mean-square error for convex combinations of gradient- and Hessian-based algorithms, assuming a random-walk model for the parameter variations. The proposed model is valid for algorithms of the same or different families, and for supervised (LMS and RLS) or blind (CMA and SWA) algorithms.

Artificial neural networks for automatic modelling of the pectus excavatum corrective prosthesis

Pectus excavatum is the most common deformity of the thorax and usually comprises Computed Tomography (CT) examination for pre-operative diagnosis. Aiming at the elimination of the high amounts of CT radiation exposure, this work presents a new methodology for the replacement of CT by a laser scanner (radiation-free) in the treatment of pectus excavatum using personally modeled prosthesis. The complete elimination of CT involves the determination of ribs external outline, at the maximum sternum depression point for prosthesis placement, based on chest wall skin surface information, acquired by a laser scanner. The developed solution resorts to artificial neural networks trained with data vectors from 165 patients. Scaled Conjugate Gradient, Levenberg-Marquardt, Resilient Back propagation and One Step Secant gradient learning algorithms were used. The training procedure was performed using the soft tissue thicknesses, determined using image processing techniques that automatically segment the skin and rib cage. The developed solution was then used to determine the ribs outline in data from 20 patient scanners. Tests revealed that ribs position can be estimated with an average error of about 6.82±5.7 mm for the left and right side of the patient. Such an error range is well below current prosthesis manual modeling (11.7±4.01 mm) even without CT imagiology, indicating a considerable step forward towards CT replacement by a 3D scanner for prosthesis personalization.

Using artificial neural networks to generate trading signals for crude oil, copper and gold futures

In this thesis, a feed-forward, back-propagating Artificial Neural Network using the gradient descent algorithm is developed to forecast the directional movement of daily returns for WTI, gold and copper futures. Out-of-sample back-test results vary, with some predictive abilities for copper futures but none for either WTI or gold. The best statistically significant hit rate achieved was 57% for copper with an absolute return Sharpe Ratio of 1.25 and a benchmarked Information Ratio of 2.11.