In this thesis, we develop tractable relaxations and efficient algorithms for large-scale optimization. Our developments are motivated by a recent paradigm, Compressed Sensing (CS), which consists of acquiring directly low-dimensional linear projections of signals, possibly corrupted with noise, and then using sophisticated recovery procedures for signal reconstruction. We start by analyzing how to utilize a priori information given in the form of sign restrictions on part of the entries. We propose necessary and sufficient on the sensing matrix for exact recovery of sparse signals, utilize them in deriving error bounds under imperfect conditions, suggest verifiable sufficient conditions and establish their limits of performance. In the second part of this thesis, we study the CS synthesis problem -selecting the minimum number of rows from a given matrix, so that the resulting submatrix possesses certifiably good recovery properties. We express the synthesis problem as the problem of approximating a given matrix by a matrix of specified low rank in the uniform norm and develop a randomized algorithm for this problem. The third part is dedicated to efficient First-Order Methods (FOMs) for large-scale, well-structured convex optimization problems. We propose FOMs with stochastic oracles that come with exact guarantees on solution quality, achieve sublinear time behavior, and through extensive simulations, show considerable improvement over the state-of-the-art deterministic FOMs. In the last part, we examine a general sparse estimation problem -estimating a block sparse linear transform of a signal from the undersampled observations of the signal corrupted with nuisance and stochastic noise. We show that an extension of the earlier results to this more general framework is possible. In particular, we suggest estimators that have efficiently verifiable guaranties of performance and provide connections to well-known results in CS theory.

<p>Time-dependent density functional theory calculations have been carried out to determine the complex first-order hyperpolarizability in the two-photon resonance region of the molecule IDS-Cab. Calculations show that three strongly absorbing states, in the ultraviolet region, are separated to the extent that no significant interference of the imaginary parts of the tensor elements of the first-order hyper-polarizability occurs. Consequently, and in contrast to experimental findings [27], no reduced imaginary parts of the first-order hyperpolarizability in the two-photon resonant region can be seen.</p>

Με σκοπό την ενίσχυση του μαθήματος Τεχνητή Νοημοσύνη στο τμήμα Μηχανικών Η/Υ και Πληροφορικής της Πολυτεχνικής σχολής του Πανεπιστημίου Πατρών έχει δημιουργηθεί από την Ομάδα Τεχνητής Νοημοσύνης το πρωτότυπο για ένα Ευφυές Σύστημα Διδασκαλίας Τεχνητής Νοημοσύνης (ΣΔΤΝ). Το σύστημα αυτό, ανάμεσα στα άλλα, διδάσκει την Κατηγορηματική Λογική ως γλώσσα Αναπαράστασης Γνώσης και Αυτόματου Συλλογισμού. Πρόκειται για ένα σύστημα που προσαρμόζεται, επιτρέποντας στους φοιτητές να επιλέγουν οι ίδιοι τον ρυθμό και το επίπεδο μάθησης. Ένα από τα θέματα που διαπραγματεύεται το σύστημα είναι και η μετατροπή από προτάσεις φυσικής γλώσσας (ΦΓ) σε εκφράσεις Κατηγορηματικού Λογισμού Πρώτης Τάξεως (ΚΛΠΤ). Επειδή η διαδικασία αυτή δεν είναι αυτοματοποιήσιμη, ο φοιτητής δεν μπορεί να πάρει κάποια βοήθεια ή υπόδειξη από το σύστημα, κατά τη διάρκεια μιας τέτοιας άσκησης, πριν δώσει την τελική του απάντηση. Γι’ αυτό, στα πλαίσια του ΣΔΤΝ αποφασίστηκε να ενσωματωθεί μια επιπλέον δυνατότητα: να μετατρέπει εκφράσεις ΚΛΠΤ τις οποίες δημιουργεί ο φοιτητής, στην προσπάθειά του να λύσει μια τέτοια άσκηση, σε προτάσεις ΦΓ. Σκοπός της λειτουργίας αυτής είναι να χρησιμοποιηθεί σαν ανατροφοδότηση από το σύστημα στον φοιτητή, προκειμένου ο τελευταίος να αξιολογήσει την απάντησή του, πριν την καταθέσει σαν τελική απάντησή στην άσκηση. Για την υλοποίηση της παραπάνω δυνατότητας ξεκίνησε η ανάπτυξη ενός συστήματος βασισμένου σε κανόνες, του FOLtoNL (First Order Logic to Natural Language). Στόχος του συστήματος ήταν η επιτυχής μετατροπή εκφράσεων ΚΛΠΤ σε ΦΓ. Το FOLtoNL υλοποιήθηκε σε Jess, μια γλώσσα προγραμματισμού με κανόνες (γραμμένη εξ’ ολοκλήρου σε Java) και αξιολογήθηκε με βάση τα αποτελέσματά του σε ειδικά σχεδιασμένο σύνολο εκφράσεων ΚΛΠΤ. / To help teaching the course of Artificial Intelligence in Computer Engineering and Informatics Department of Patras University, a web-based intelligent tutoring system, called Artificial Intelligence Teaching System (AITS), was created. Among other things, AITS teaches Predicate Logic as a Knowledge Representation and Automated Reasoning language and is an adapting system, allowing students to choose themselves the teaching rate and level. One of the issues that AITS deals with is the conversion of natural language (NL) sentences into First-Order Logic (FOL) formulas. Given that this is a non-automated process, it is difficult to give some hints to the students-users during their effort to convert an “unknown” (to the system) NL sentence into a FOL formula. However, some kind of help could be provided, if the system could translate (after checking its syntax) the proposed by the student FOL formula into a NL sentence. The student then will be able to compare the initial NL sentence with the one that its FOL formula corresponds to. In this way, it is easier to see whether his/her proposed FOL formula is compatible with the given NL sentence and perhaps make some amendments, before submitting the final answer. FOLtoNL (First Order Logic to Natural Language) is a rule-based system that converts FOL formulas into NL in order to provide the functionality described above. It uses the expert systems approach alongside natural language processing aspects. FOLtoNL is implemented in Jess (an expert system shell written in Java) and has been evaluated via an appropriately created set of FOL expressions.

In a series of work initiated by Nemirovsky and Yudin, and later extended by Nesterov, first-order algorithms for unconstrained minimization with optimal theoretical complexity bound have been proposed. On the other hand, conjugate gradient algorithms as one of the widely used first-order techniques suffer from the lack of a finite complexity bound. In fact their performance can possibly be quite poor. This dissertation is partially on tightening the gap between these two classes of algorithms, namely the traditional conjugate gradient methods and optimal first-order techniques. We derive conditions under which conjugate gradient methods attain the same complexity bound as in Nemirovsky-Yudin's and Nesterov's methods. Moreover, we propose a conjugate gradient-type algorithm named CGSO, for Conjugate Gradient with Subspace Optimization, achieving the optimal complexity bound with the payoff of a little extra computational cost. We extend the theory of CGSO to convex problems with linear constraints. In particular we focus on solving $l_1$-regularized least square problem, often referred to as Basis Pursuit Denoising (BPDN) problem in the optimization community. BPDN arises in many practical fields including sparse signal recovery, machine learning, and statistics. Solving BPDN is fairly challenging because the size of the involved signals can be quite large; therefore first order methods are of particular interest for these problems. We propose a quasi-Newton proximal method for solving BPDN. Our numerical results suggest that our technique is computationally effective, and can compete favourably with the other state-of-the-art solvers.

Indirect reciprocity describes a class of reputation-based mechanisms which may explain the prevalence of cooperation in groups where partners meet only once. The first model for which this has analytically been shown was the binary image scoring mechanism, where one's reputation is only based on one's last action. But this mechanism is known to fail if errors in implementation occur. It has thus been claimed that for indirect reciprocity to stabilize cooperation, reputation assessments must be of higher order, i.e. contingent not only on past actions, but also on the reputations of the targets of these actions. We show here that this need not be the case. A simple image scoring mechanism where more than just one past action is observed provides ample possibilities for stable cooperation to emerge even under substantial rates of implementation errors. (authors' abstract) / Series: Department of Economics Working Paper Series

During the last decade, the use of time scales as a means of unifying and extending results about various types of dynamic equations has proven to be both prolific and fruitful. Many classical results from the theories of differential and difference equations have time scale analogues. In this thesis we derive new oscillation criteria for second order dynamic equations on time scales.

Les réseaux biologiques, notamment les réseaux de signalisation déclenchés par les hormones, sont extrêmement complexes. Les méthodes expérimentales à haut débit permettent d’aborder cette complexité, mais la prise en compte de l’ensemble des données générées requiert la mise au point de méthodes automatiques pour la construction des réseaux. Nous avons développé une nouvelle méthode d’inférence reposant sur la formalisation, sous forme de règles logiques, du raisonnement de l’expert sur les données expérimentales. Cela nécessite la constitution d’une base de connaissances, ensuite exploitée par un moteur d’inférence afin de déduire les conclusions permettant de construire les réseaux. Notre méthode a été élaborée grâce au réseau de signalisation induit par l’hormone folliculo-stimulante dont le récepteur fait partie de la grande famille des récepteurs couplés aux protéines G. Ce réseau a également été construit manuellement pour évaluer notre méthode. Un contrôle a ensuite été réalisé sur réseau induit par le facteur de croissance épidermique, se liant à un récepteur tyrosine kinase, de façon à montrer que notre méthode est capable de déduire différents types de réseaux de signalisation. / Biological networks, including signalling networks induced by hormones, are very complex. High-throughput experimental methods permit to approach this complexity, but to be able to use all generated data, it is necessary to create automatical inference methods to build networks. We have developped a new inference method based on the formalization of the expert’s reasoning on experimental data. This reasoning is converted into logical rules. This work requires the creation of a knowledge base which is used by an inference engine to deduce conclusions to build networks. Our method has been elaborated by the construction of the signalling network induced by the follicle stimulating hormone whose receptor belongs to the G protein-coupled receptors family. This network has also been built manually to assess our method. Then, a test has been done on the network induced by the epidermal growth factor, which binds to a tyrosine kinase receptor, to demonstrate the ability of our method to deduce differents types of signaling networks.

Gramática de grafos é uma linguagem formal bastante adequada para sistemas cujos estados possuem uma topologia complexa (que envolvem vários tipos de elementos e diferentes tipos de relações entre eles) e cujo comportamento é essencialmente orientado pelos dados, isto é, eventos são disparados por configurações particulares do estado. Vários sistemas reativos são exemplos desta classe de aplicações, como protocolos para sistemas distribuídos e móveis, simulação de sistemas biológicos, entre outros. A verificação de gramática de grafos através da técnica de verificação de modelos já é utilizada por diversas abordagens. Embora esta técnica constitua um método de análise bastante importante, ela tem como desvantagem a necessidade de construir o espaço de estados completo do sistema, o que pode levar ao problema da explosão de estados. Bastante progresso tem sido feito para lidar com esta dificuldade, e diversas técnicas têm aumentado o tamanho dos sistemas que podem ser verificados. Outras abordagens propõem aproximar o espaço de estados, mas neste caso não é possível a verificação de propriedades arbitrárias. Além da verificação de modelos, a prova de teoremas constitui outra técnica consolidada para verificação formal. Nesta técnica tanto o sistema quanto suas propriedades são expressas em alguma lógica matemática. O processo de prova consiste em encontrar uma prova a partir dos axiomas e lemas intermediários do sistema. Cada técnica tem argumentos pró e contra o seu uso, mas é possível dizer que a verificação de modelos e a prova de teoremas são complementares. A maioria das abordagens utilizam verificadores de modelos para analisar propriedades de computações, isto é, sobre a seqüência de passos de um sistema. Propriedades sobre estados alcançáveis só são verificadas de forma restrita. O objetivo deste trabalho é prover uma abordagem para a prova de propriedades de grafos alcançáveis de uma gramática de grafos através da técnica de prova de teoremas. Propõe-se uma tradução (da abordagem Single-Pushout) de gramática de grafos para uma abordagem lógica e relacional, a qual permite a aplicação de indução matemática para análise de sistemas com espaço de estados infinito. Definiu-se gramática de grafos utilizando estruturas relacionais e aplicações de regras com linguagens lógicas. Inicialmente considerou-se o caso de grafos (tipados) simples, e então se estendeu a abordagem para grafos com atributos e gramáticas com condições negativas de aplicação. Além disso, baseado nesta abordagem, foram estabelecidos padrões para a definição, codificação e reuso de especificações de propriedades. O sistema de padrões tem o objetivo de auxiliar e simplificar a tarefa de especificar requisitos de forma precisa. Finalmente, propõe-se implementar definições relacionais de gramática de grafos em estruturas de event-B, de forma que seja possível utilizar os provadores disponíveis para event-B para demonstrar propriedades de gramática de grafos. / Graph grammars are a formal language well-suited to applications in which states have a complex topology (involving not only many types of elements, but also different types of relations between them) and in which behaviour is essentially data-driven, that is, events are triggered basically by particular configurations of the state. Many reactive systems are examples of this class of applications, such as protocols for distributed and mobile systems, simulation of biological systems, and many others. The verification of graph grammar models through model-checking is currently supported by various approaches. Although model-checking is an important analysis method, it has as disadvantage the need to build the complete state space, which can lead to the state explosion problem. Much progress has been made to deal with this difficulty, and many techniques have increased the size of the systems that may be verified. Other approaches propose to over- and/or under-approximate the state-space, but in this case it is not possible to check arbitrary properties. Besides model checking, theorem proving is another wellestablished approach for verification. Theorem proving is a technique where both the system and its desired properties are expressed as formulas in some mathematical logic. A logical description defines the system, establishing a set of axioms and inference rules. The process of verification consists of finding a proof of the required property from the axioms or intermediary lemmas of the system. Each verification technique has arguments for and against its use, but we can say that model-checking and theorem proving are complementary. Most of the existing approaches use model checkers to analyse properties of computations, that is, properties over the sequences of steps a system may engage in. Properties about reachable states are handled, if at all possible, only in very restricted ways. In this work, our main aim is to provide a means to prove properties of reachable graphs of graph grammar models using the theorem proving technique. We propose an encoding of (the Single-Pushout approach of) graph grammar specifications into a relational and logical approach which allows the application of the mathematical induction technique to analyse systems with infinite state-spaces. We have defined graph grammars using relational structures and used logical languages to model rule applications. We first consider the case of simple (typed) graphs, and then we extend the approach to the non-trivial case of attributed-graphs and grammars with negative application conditions. Besides that, based on this relational encoding, we establish patterns for the presentation, codification and reuse of property specifications. The pattern has the goal of helping and simplifying the task of stating precise requirements to be verified. Finally, we propose to implement relational definitions of graph grammars in event-B structures, such that it is possible to use the event-B provers to demonstrate properties of a graph grammar.

Gramática de grafos é uma linguagem formal bastante adequada para sistemas cujos estados possuem uma topologia complexa (que envolvem vários tipos de elementos e diferentes tipos de relações entre eles) e cujo comportamento é essencialmente orientado pelos dados, isto é, eventos são disparados por configurações particulares do estado. Vários sistemas reativos são exemplos desta classe de aplicações, como protocolos para sistemas distribuídos e móveis, simulação de sistemas biológicos, entre outros. A verificação de gramática de grafos através da técnica de verificação de modelos já é utilizada por diversas abordagens. Embora esta técnica constitua um método de análise bastante importante, ela tem como desvantagem a necessidade de construir o espaço de estados completo do sistema, o que pode levar ao problema da explosão de estados. Bastante progresso tem sido feito para lidar com esta dificuldade, e diversas técnicas têm aumentado o tamanho dos sistemas que podem ser verificados. Outras abordagens propõem aproximar o espaço de estados, mas neste caso não é possível a verificação de propriedades arbitrárias. Além da verificação de modelos, a prova de teoremas constitui outra técnica consolidada para verificação formal. Nesta técnica tanto o sistema quanto suas propriedades são expressas em alguma lógica matemática. O processo de prova consiste em encontrar uma prova a partir dos axiomas e lemas intermediários do sistema. Cada técnica tem argumentos pró e contra o seu uso, mas é possível dizer que a verificação de modelos e a prova de teoremas são complementares. A maioria das abordagens utilizam verificadores de modelos para analisar propriedades de computações, isto é, sobre a seqüência de passos de um sistema. Propriedades sobre estados alcançáveis só são verificadas de forma restrita. O objetivo deste trabalho é prover uma abordagem para a prova de propriedades de grafos alcançáveis de uma gramática de grafos através da técnica de prova de teoremas. Propõe-se uma tradução (da abordagem Single-Pushout) de gramática de grafos para uma abordagem lógica e relacional, a qual permite a aplicação de indução matemática para análise de sistemas com espaço de estados infinito. Definiu-se gramática de grafos utilizando estruturas relacionais e aplicações de regras com linguagens lógicas. Inicialmente considerou-se o caso de grafos (tipados) simples, e então se estendeu a abordagem para grafos com atributos e gramáticas com condições negativas de aplicação. Além disso, baseado nesta abordagem, foram estabelecidos padrões para a definição, codificação e reuso de especificações de propriedades. O sistema de padrões tem o objetivo de auxiliar e simplificar a tarefa de especificar requisitos de forma precisa. Finalmente, propõe-se implementar definições relacionais de gramática de grafos em estruturas de event-B, de forma que seja possível utilizar os provadores disponíveis para event-B para demonstrar propriedades de gramática de grafos. / Graph grammars are a formal language well-suited to applications in which states have a complex topology (involving not only many types of elements, but also different types of relations between them) and in which behaviour is essentially data-driven, that is, events are triggered basically by particular configurations of the state. Many reactive systems are examples of this class of applications, such as protocols for distributed and mobile systems, simulation of biological systems, and many others. The verification of graph grammar models through model-checking is currently supported by various approaches. Although model-checking is an important analysis method, it has as disadvantage the need to build the complete state space, which can lead to the state explosion problem. Much progress has been made to deal with this difficulty, and many techniques have increased the size of the systems that may be verified. Other approaches propose to over- and/or under-approximate the state-space, but in this case it is not possible to check arbitrary properties. Besides model checking, theorem proving is another wellestablished approach for verification. Theorem proving is a technique where both the system and its desired properties are expressed as formulas in some mathematical logic. A logical description defines the system, establishing a set of axioms and inference rules. The process of verification consists of finding a proof of the required property from the axioms or intermediary lemmas of the system. Each verification technique has arguments for and against its use, but we can say that model-checking and theorem proving are complementary. Most of the existing approaches use model checkers to analyse properties of computations, that is, properties over the sequences of steps a system may engage in. Properties about reachable states are handled, if at all possible, only in very restricted ways. In this work, our main aim is to provide a means to prove properties of reachable graphs of graph grammar models using the theorem proving technique. We propose an encoding of (the Single-Pushout approach of) graph grammar specifications into a relational and logical approach which allows the application of the mathematical induction technique to analyse systems with infinite state-spaces. We have defined graph grammars using relational structures and used logical languages to model rule applications. We first consider the case of simple (typed) graphs, and then we extend the approach to the non-trivial case of attributed-graphs and grammars with negative application conditions. Besides that, based on this relational encoding, we establish patterns for the presentation, codification and reuse of property specifications. The pattern has the goal of helping and simplifying the task of stating precise requirements to be verified. Finally, we propose to implement relational definitions of graph grammars in event-B structures, such that it is possible to use the event-B provers to demonstrate properties of a graph grammar.

Made available in DSpace on 2015-05-14T12:14:06Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 1801004 bytes, checksum: f4e9da048482e04e21d6f5a9ffe45a4f (MD5) Previous issue date: 2006-02-22 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / We start on work with a brief revision of gravitation and cosmology. We use the standard cosmological model of Friedmann-Robertson-Walker to investigate cosmological models in the presence of real scalar fields. The subject is of interest to dark energy. As main result of this work, we show that the equations of motion can be written as first-order differential equations. In general much simpler to be solved by standard methods. / Iniciamos o nosso trabalho com uma rápida revisão de cosmologia e gravitação. Utilizamos o modelo cosmológico padrão, a métrica de Friedmann-Robertson-Walker para estudarmos modelos cosmológicos na presença de um campos escalares reais. O assunto ´e de interesse para a energia escura. Como resultado principal deste trabalho nós mostramos que as equações de movimento podem ser escritas como equações diferenciais de primeira ordem. Em geral muito mais simples de serem resolvidos pelos métodos usuais.