Advances in scaling deep learning algorithms

Les algorithmes d'apprentissage profond forment un nouvel ensemble de méthodes puissantes pour l'apprentissage automatique. L'idée est de combiner des couches de facteurs latents en hierarchies. Cela requiert souvent un coût computationel plus elevé et augmente aussi le nombre de paramètres du modèle. Ainsi, l'utilisation de ces méthodes sur des problèmes à plus grande échelle demande de réduire leur coût et aussi d'améliorer leur régularisation et leur optimization. Cette thèse adresse cette question sur ces trois perspectives. Nous étudions tout d'abord le problème de réduire le coût de certains algorithmes profonds. Nous proposons deux méthodes pour entrainer des machines de Boltzmann restreintes et des auto-encodeurs débruitants sur des distributions sparses à haute dimension. Ceci est important pour l'application de ces algorithmes pour le traitement de langues naturelles. Ces deux méthodes (Dauphin et al., 2011; Dauphin and Bengio, 2013) utilisent l'échantillonage par importance pour échantilloner l'objectif de ces modèles. Nous observons que cela réduit significativement le temps d'entrainement. L'accéleration atteint 2 ordres de magnitude sur plusieurs bancs d'essai. Deuxièmement, nous introduisont un puissant régularisateur pour les méthodes profondes. Les résultats expérimentaux démontrent qu'un bon régularisateur est crucial pour obtenir de bonnes performances avec des gros réseaux (Hinton et al., 2012). Dans Rifai et al. (2011), nous proposons un nouveau régularisateur qui combine l'apprentissage non-supervisé et la propagation de tangente (Simard et al., 1992). Cette méthode exploite des principes géometriques et permit au moment de la publication d'atteindre des résultats à l'état de l'art. Finalement, nous considérons le problème d'optimiser des surfaces non-convexes à haute dimensionalité comme celle des réseaux de neurones. Tradionellement, l'abondance de minimum locaux était considéré comme la principale difficulté dans ces problèmes. Dans Dauphin et al. (2014a) nous argumentons à partir de résultats en statistique physique, de la théorie des matrices aléatoires, de la théorie des réseaux de neurones et à partir de résultats expérimentaux qu'une difficulté plus profonde provient de la prolifération de points-selle. Dans ce papier nous proposons aussi une nouvelle méthode pour l'optimisation non-convexe.

Computationally reliable approaches of contractive MPC for discrete-time systems

La tesis pretende explorar acercamientos computacionalmente confiables y eficientes de contractivo MPC para sistemas de tiempo discreto. Dos tipos de contractivo MPC han sido estudiados: MPC con coacción contractiva obligatoria y MPC con una secuencia contractiva de conjuntos controlables. Las técnicas basadas en optimización convexa y análisis de intervalos son aplicadas para tratar MPC contractivo lineal y no lineal, respectivamente. El análisis de intervalos clásicos es ampliado a zonotopes en la geometría para diseñar un conjunto invariante de control terminal para el modo dual de MPC. También es ampliado a intervalos modales para tener en cuenta la modalidad al calcula de conjuntos controlables robustos con una interpretación semántica clara. Los instrumentos de optimización convexa y análisis de intervalos han sido combinados para mejorar la eficacia de contractive MPC para varias clases de sistemas de tiempo discreto inciertos no lineales limitados. Finalmente, los dos tipos dirigidos de contractivo MPC han sido aplicados para controlar un Torneo de Fútbol de Copa Mundial de Micro Robot (MiroSot) y un Tanque-Reactor de Mezcla Continua (CSTR), respectivamente.

Controle preditivo robusto baseado em desigualdades matriciais lineares aplicado a um sistema de tanques acoplados

This work deals with an on-line control strategy based on Robust Model Predictive Control (RMPC) technique applied in a real coupled tanks system. This process consists of two coupled tanks and a pump to feed the liquid to the system. The control objective (regulator problem) is to keep the tanks levels in the considered operation point even in the presence of disturbance. The RMPC is a technique that allows explicit incorporation of the plant uncertainty in the problem formulation. The goal is to design, at each time step, a state-feedback control law that minimizes a 'worst-case' infinite horizon objective function, subject to constraint in the control. The existence of a feedback control law satisfying the input constraints is reduced to a convex optimization over linear matrix inequalities (LMIs) problem. It is shown in this work that for the plant uncertainty described by the polytope, the feasible receding horizon state feedback control design is robustly stabilizing. The software implementation of the RMPC is made using Scilab, and its communication with Coupled Tanks Systems is done through the OLE for Process Control (OPC) industrial protocol

H-2 and H-infinity - norm control of intelligent structures using LMI techniques

Linear Matrix Inequalities (LMIs) is a powerful too] that has been used in many areas ranging from control engineering to system identification and structural design. There are many factors that make LMI appealing. One is the fact that a lot of design specifications and constrains can be formulated as LMIs [1]. Once formulated in terms of LMIs a problem can be solved efficiently by convex optimization algorithms. The basic idea of the LMI method is to formulate a given problem as an optimization problem with linear objective function and linear matrix inequalities constrains. An intelligent structure involves distributed sensors and actuators and a control law to apply localized actions, in order to minimize or reduce the response at selected conditions. The objective of this work is to implement techniques of control based on LMIs applied to smart structures.

Robust control applications for smart truss structure

A finite element modeling of an intelligent truss structure with piezoelectric stack actuators for the purpose of active damping and structural vibration attenuation is presented. This paper concerns with the following issues aspects: the design of intelligent truss structure considering electro-mechanical coupling between the host structure and piezoelectric stack actuators; the H 2 norm approach to search for optimal placement of actuators and sensors; and finally some aspects in robust control techniques. The electro-mechanical behavior of piezoelectric elements is directly related to the successful application of the actuators in truss structures. In order to achieve the desired damping in the interested bandwidth frequency it is used the H ∞ output feedback solved by convex optimization. The constraints to be reached are written by linear matrix inequalities (LMI). The paper concludes with a numerical example, using Matlab and Simulink, in a cantilevered, 2-bay space truss structure. The results demonstrated the approach applicability.

Active control in flexible plates with piezoelectric actuators using linear matrix inequalities

The study of algorithms for active vibrations control in flexible structures became an area of enormous interest, mainly due to the countless demands of an optimal performance of mechanical systems as aircraft, aerospace and automotive structures. Smart structures, formed by a structure base, coupled with piezoelectric actuators and sensor are capable to guarantee the conditions demanded through the application of several types of controllers. The actuator/sensor materials are composed by piezoelectric ceramic (PZT - Lead Zirconate Titanate), commonly used as distributed actuators, and piezoelectric plastic films (PVDF-PolyVinyliDeno Floride), highly indicated for distributed sensors. The design process of such system encompasses three main phases: structural design; optimal placement of sensor/actuator (PVDF and PZT); and controller design. Consequently, for optimal design purposes, the structure, the sensor/actuator placement and the controller have to be considered simultaneously. This article addresses the optimal placement of actuators and sensors for design of controller for vibration attenuation in a flexible plate. Techniques involving linear matrix inequalities (LMI) to solve the Riccati's equation are used. The controller's gain is calculated using the linear quadratic regulator (LQR). The major advantage of LMI design is to enable specifications such as stability degree requirements, decay rate, input force limitation in the actuators and output peak bounder. It is also possible to assume that the model parameters involve uncertainties. LMI is a very useful tool for problems with constraints, where the parameters vary in a range of values. Once formulated in terms of LMI a problem can be solved efficiently by convex optimization algorithms.

An application of a modified constructive heuristic algorithm to transmission expansion planning

Transmission expansion planning (TEP) is a non-convex optimization problem that can be solved via different heuristic algorithms. A variety of classical as well as heuristic algorithms in literature are addressed to solve TEP problem. In this paper a modified constructive heuristic algorithm (CHA) is proposed for solving such a crucial problem. Most of research papers handle TEP problem by linearization of the non-linear mathematical model while in this research TEP problem is solved via CHA using non-linear model. The proposed methodology is based upon Garver's algorithm capable of applying to a DC model. Simulation studies and tests results on the well known transmission network such as: Garver and IEEE 24-bus systems are carried out to show the significant performance as well as the effectiveness of the proposed algorithm. © 2011 IEEE.

An application of CHA to concurrent short-term transmission expansion & reactive power planning

In this paper a heuristic technique for solving simultaneous short-term transmission network expansion and reactive power planning problem (TEPRPP) via an AC model is presented. A constructive heuristic algorithm (CHA) aimed to obtaining a significant quality solution for such problem is employed. An interior point method (IPM) is applied to solve TEPRPP as a nonlinear programming (NLP) during the solution steps of the algorithm. For each proposed network topology, an indicator is deployed to identify the weak buses for reactive power sources placement. The objective function of NLP includes the costs of new transmission lines, real power losses as well as reactive power sources. By allocating reactive power sources at load buses, the circuit capacity may increase while the cost of new lines can be decreased. The proposed methodology is tested on Garver's system and the obtained results shows its capability and the viability of using AC model for solving such non-convex optimization problem. © 2011 IEEE.

Identificação experimental e controle ativo de vibrações aplicadas em estruturas inteligentes

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Projeto de controladores robustos para aplicações em estruturas inteligentes utilizando desigualdades matriciais lineares

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Controle ativo de vibrações em estruturas flexíveis utilizando desigualdades matriciais lineares (LMIs)

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Modificação de zeros em sistemas de controle robusto utilizando LMIs

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Aeroelastic stability analysis using linear matrix inequalities

The present work describes an alternative methodology for identification of aeroelastic stability in a range of varying parameters. Analysis is performed in time domain based on Lyapunov stability and solved by convex optimization algorithms. The theory is outlined and simulations are carried out on a benchmark system to illustrate the method. The classical methodology with the analysis of the system's eigenvalues is presented for comparing the results and validating the approach. The aeroelastic model is represented in state space format and the unsteady aerodynamic forces are written in time domain using rational function approximation. The problem is formulated as a polytopic differential inclusion system and the conceptual idea can be used in two different applications. In the first application the method verifies the aeroelastic stability in a range of air density (or its equivalent altitude range). In the second one, the stability is verified for a rage of velocities. These analyses are in contrast to the classical discrete analysis performed at fixed air density/velocity values. It is shown that this method is efficient to identify stability regions in the flight envelope and it offers promise for robust flutter identification.

Control of Limit Cycle Oscillation in a Three Degrees of Freedom Airfoil Section Using Fuzzy Takagi-Sugeno Modeling

This work presents a strategy to control nonlinear responses of aeroelastic systems with control surface freeplay. The proposed methodology is developed for the three degrees of freedom typical section airfoil considering aerodynamic forces from Theodorsen's theory. The mathematical model is written in the state space representation using rational function approximation to write the aerodynamic forces in time domain. The control system is designed using the fuzzy Takagi-Sugeno modeling to compute a feedback control gain. It useds Lyapunov's stability function and linear matrix inequalities (LMIs) to solve a convex optimization problem. Time simulations with different initial conditions are performed using a modified Runge-Kutta algorithm to compare the system with and without control forces. It is shown that this approach can compute linear control gain able to stabilize aeroelastic systems with discontinuous nonlinearities.

Planejamento de reativos em sistemas elétricos de potência multi-área através de modelos estocásticos

Pós-graduação em Engenharia Elétrica - FEIS