Justified Generalization: Acquiring Procedures from Examples

This thesis describes an implemented system called NODDY for acquiring procedures from examples presented by a teacher. Acquiring procedures form examples involves several different generalization tasks. Generalization is an underconstrained task, and the main issue of machine learning is how to deal with this underconstraint. The thesis presents two principles for constraining generalization on which NODDY is based. The first principle is to exploit domain based constraints. NODDY demonstrated how such constraints can be used both to reduce the space of possible generalizations to manageable size, and how to generate negative examples out of positive examples to further constrain the generalization. The second principle is to avoid spurious generalizations by requiring justification before adopting a generalization. NODDY demonstrates several different ways of justifying a generalization and proposes a way of ordering and searching a space of candidate generalizations based on how much evidence would be required to justify each generalization. Acquiring procedures also involves three types of constructive generalizations: inferring loops (a kind of group), inferring complex relations and state variables, and inferring predicates. NODDY demonstrates three constructive generalization methods for these kinds of generalization.

Causality enabled compositional modelling of Bayesian networks

J. Keppens and Q. Shen. Causality enabled compositional modelling of Bayesian networks. Proceedings of the 18th International Workshop on Qualitative Reasoning, pages 33-40, 2004.

Cooperation between adsorbates accounts for the activation of atomic layer deposition reactions

Atomic layer deposition (ALD) is a technique for producing conformal layers of nanometre-scale thickness, used commercially in non-planar electronics and increasingly in other high-tech industries. ALD depends on self-limiting surface chemistry but the mechanistic reasons for this are not understood in detail. Here we demonstrate, by first-principle calculations of growth of HfO2 from Hf(N(CH3)2)4–H2O and HfCl4–H2O and growth of Al2O3 from Al(CH3)3–H2O, that, for all these precursors, co-adsorption plays an important role in ALD. By this we mean that previously-inert adsorbed fragments can become reactive once sufficient numbers of molecules adsorb in their neighbourhood during either precursor pulse. Through the calculated activation energies, this ‘cooperative’ mechanism is shown to have a profound influence on proton transfer and ligand desorption, which are crucial steps in the ALD cycle. Depletion of reactive species and increasing coordination cause these reactions to self-limit during one precursor pulse, but to be re-activated via the cooperative effect in the next pulse. This explains the self-limiting nature of ALD.

Beyond simple imaging with energetic beams – nanopatterning, correlative microscopy and statistical analysis of inclusions

Electron microscopy (EM) has advanced in an exponential way since the first transmission electron microscope (TEM) was built in the 1930’s. The urge to ‘see’ things is an essential part of human nature (talk of ‘seeing is believing’) and apart from scanning tunnel microscopes which give information about the surface, EM is the only imaging technology capable of really visualising atomic structures in depth down to single atoms. With the development of nanotechnology the demand to image and analyse small things has become even greater and electron microscopes have found their way from highly delicate and sophisticated research grade instruments to key-turn and even bench-top instruments for everyday use in every materials research lab on the planet. The semiconductor industry is as dependent on the use of EM as life sciences and pharmaceutical industry. With this generalisation of use for imaging, the need to deploy advanced uses of EM has become more and more apparent. The combination of several coinciding beams (electron, ion and even light) to create DualBeam or TripleBeam instruments for instance enhances the usefulness from pure imaging to manipulating on the nanoscale. And when it comes to the analytic power of EM with the many ways the highly energetic electrons and ions interact with the matter in the specimen there is a plethora of niches which evolved during the last two decades, specialising in every kind of analysis that can be thought of and combined with EM. In the course of this study the emphasis was placed on the application of these advanced analytical EM techniques in the context of multiscale and multimodal microscopy – multiscale meaning across length scales from micrometres or larger to nanometres, multimodal meaning numerous techniques applied to the same sample volume in a correlative manner. In order to demonstrate the breadth and potential of the multiscale and multimodal concept an integration of it was attempted in two areas: I) Biocompatible materials using polycrystalline stainless steel and II) Semiconductors using thin multiferroic films. I) The motivation to use stainless steel (316L medical grade) comes from the potential modulation of endothelial cell growth which can have a big impact on the improvement of cardio-vascular stents – which are mainly made of 316L – through nano-texturing of the stent surface by focused ion beam (FIB) lithography. Patterning with FIB has never been reported before in connection with stents and cell growth and in order to gain a better understanding of the beam-substrate interaction during patterning a correlative microscopy approach was used to illuminate the patterning process from many possible angles. Electron backscattering diffraction (EBSD) was used to analyse the crystallographic structure, FIB was used for the patterning and simultaneously visualising the crystal structure as part of the monitoring process, scanning electron microscopy (SEM) and atomic force microscopy (AFM) were employed to analyse the topography and the final step being 3D visualisation through serial FIB/SEM sectioning. II) The motivation for the use of thin multiferroic films stems from the ever-growing demand for increased data storage at lesser and lesser energy consumption. The Aurivillius phase material used in this study has a high potential in this area. Yet it is necessary to show clearly that the film is really multiferroic and no second phase inclusions are present even at very low concentrations – ~0.1vol% could already be problematic. Thus, in this study a technique was developed to analyse ultra-low density inclusions in thin multiferroic films down to concentrations of 0.01%. The goal achieved was a complete structural and compositional analysis of the films which required identification of second phase inclusions (through elemental analysis EDX(Energy Dispersive X-ray)), localise them (employing 72 hour EDX mapping in the SEM), isolate them for the TEM (using FIB) and give an upper confidence limit of 99.5% to the influence of the inclusions on the magnetic behaviour of the main phase (statistical analysis).

Enhancing process insight in polymer extrusion by grey box modelling

Due to the complexity and inherent instability in polymer extrusion there is a need for process models which can be run on-line to optimise settings and control disturbances. First-principle models demand computationally intensive solution, while ‘black box’ models lack generalisation ability and physical process insight. This work examines a novel ‘grey box’ modelling technique which incorporates both prior physical knowledge and empirical data in generating intuitive models of the process. The models can be related to the underlying physical mechanisms in the extruder and have been shown to capture unpredictable effects of the operating conditions on process instability. Furthermore, model parameters can be related to material properties available from laboratory analysis and as such, lend themselves to re-tuning for different materials without extensive remodelling work.

Electrostatic solitary waves in the presence of excess superthermal electrons: modulational instability and envelope soliton modes

The nonlinear dynamics of electrostatic solitary waves in the form of localized modulated wavepackets is investigated from first principles. Electron-acoustic (EA) excitations are considered in a two-electron plasma, via a fluid formulation. The plasma, assumed to be collisionless and uniform (unmagnetized), is composed of two types of electrons (inertial cold electrons and inertialess kappa-distributed superthermal electrons) and stationary ions. By making use of a multiscale perturbation technique, a nonlinear Schrodinger equation is derived for the modulated envelope, relying on which the occurrence of modulational instability (MI) is investigated in detail. Stationary profile localized EA excitations may exist, in the form of bright solitons (envelope pulses) or dark envelopes (voids). The presence of superthermal electrons modifies the conditions for MI to occur, as well as the associated threshold and growth rate. The concentration of superthermal electrons (i.e., the deviation from a Maxwellian electron distribution) may control or even suppress MI. Furthermore, superthermality affects the characteristics of solitary envelope structures, both qualitatively (supporting one or the other type, for different.) and quantitatively, changing their characteristics (width, amplitude). The stability of bright and dark-type nonlinear structures is confirmed by numerical simulations.

A Unified Scheduler for Recursive and Task Dataflow Parallelism

Task dataflow languages simplify the specification of parallel programs by dynamically detecting and enforcing dependencies between tasks. These languages are, however, often restricted to a single level of parallelism. This language design is reflected in the runtime system, where a master thread explicitly generates a task graph and worker threads execute ready tasks and wake-up their dependents. Such an approach is incompatible with state-of-the-art schedulers such as the Cilk scheduler, that minimize the creation of idle tasks (work-first principle) and place all task creation and scheduling off the critical path. This paper proposes an extension to the Cilk scheduler in order to reconcile task dependencies with the work-first principle. We discuss the impact of task dependencies on the properties of the Cilk scheduler. Furthermore, we propose a low-overhead ticket-based technique for dependency tracking and enforcement at the object level. Our scheduler also supports renaming of objects in order to increase task-level parallelism. Renaming is implemented using versioned objects, a new type of hyper object. Experimental evaluation shows that the unified scheduler is as efficient as the Cilk scheduler when tasks have no dependencies. Moreover, the unified scheduler is more efficient than SMPSS, a particular implementation of a task dataflow language.

Rational screening low-cost counter electrodes for dye-sensitized solar cells

Dye-sensitized solar cells have attracted intense research attention owing to their ease of fabrication, cost-effectiveness and high efficiency in converting solar energy. Noble platinum is generally used as catalytic counter electrode for redox mediators in electrolyte solution. Unfortunately, platinum is expensive and non-sustainable for long-term applications. Therefore, researchers are facing with the challenge of developing low-cost and earth-abundant alternatives. So far, rational screening of non-platinum counter electrodes has been hamstrung by the lack of understanding about the electrocatalytic process of redox mediators on various counter electrodes. Here, using first-principle quantum chemical calculations, we studied the electrocatalytic process of redox mediators and predicted electrocatalytic activity of potential semiconductor counter electrodes. On the basis of theoretical predictions, we successfully used rust (alpha-Fe2O3) as a new counter electrode catalyst, which demonstrates promising electrocatalytic activity towards triiodide reduction at a rate comparable to platinum.

Productive Knowledge of English Binomials by Japanese Learners of English

The purpose of this research is to reveal (1) which English binomials Japanese learners of English have productive knowledge of and (2) what strategies they use to produce English binomials when they do not know the binomials. One hundred and three Japanese learners of English with intermediate proficiency level completed an online survey of 44 binomials. The participants were given the first word of a binomial and asked to type a word following “and”. The target word was provided by more than 75% of participants for 19 of the 44 binomials, meaning that learners have productive knowledge for certain binomials. An analysis of errors suggested that the participants relied heavily on semantic relationships between items in binomials.However, the use of a semantic strategy for producing the second words often leads to non-binomial expressions. From these results we suggest that giving more input to learners, as well as teaching the “Me first” principle (Cooper & Ross, 1975) explicitly would help the learners to develop more accurate and effective strategies for uncertain or unfamiliar binomials.

Desenvolvimento de modelo de primeiros princípios para simulação dinâmica de uma coluna de destilação contínua

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia Química e Biológica

Heróis de palmo e meio: a intervenção psicossocial e a relação de ajuda no acolhimento institucional

Os Lares de Infância e Juventude têm como objetivo defender o superior interesse das crianças, desempenhando um papel fundamental no acompanhamento do seu desenvolvimento. Este dever concretiza-se substancialmente na definição e acompanhamento dos seus projetos de vida. Neste sentido, as práticas dos profissionais devem ir ao encontro das reais necessidades dos seus atores, numa ação adequada, desenhada e concertada, tendo sempre como principio básico o respeito pelas idiossincrasias de cada criança e jovem. O presente Relatório tem como objetivo apresentar o desenho e desenvolvimento de um projeto de intervenção psicossocial e relação de ajuda com crianças em acolhimento institucional, intitulado “Heróis de Palmo e Meio. A intervenção psicossocial e a relação de ajuda no acolhimento institucional”. Este projeto teve como grande finalidade a “Participação das crianças e das suas famílias na construção dos seus projetos de vida com base na compreensão da institucionalização”. Priorizando as necessidades e respeitando o critério de urgência, foram desenhados dois subprojectos com duas crianças e suas famílias, tentando sempre responder às suas necessidades individuais. Foram definidos alguns objetivos gerais comuns, assim como objetivos específicos, estratégias e ações adequadas a cada um dos subprojectos. Os resultados deste projeto revelam-se significativos, já que permitiram, quer às crianças, quer às famílias, desenvolver um processo de reflexão e consciencialização em torno da sua história de vida, compreensão da sua realidade e participação ativa no projeto de vida dos filhos, o que veio facilitar o caminho inicial do processo da mudança.

Calculation of the optical conductivity for the Anderson impurity model using the slave boson technique

The optical conductivity of the Anderson impurity mode l has been calculated by emp l oying the slave boson technique and an expansion in powers of l i N, where N is the d egeneracy o f the f electron level . This method has been used to find the effective mass of the conduction electrons for temperatures above and below the Kondo tempera ture. For low temperatures, the mass enhancement is f ound to be large while a t high t emperatures, the mass enhancement is sma ll. The conductivity i s f ound to be Drude like with frequency dependent effective mass and scattering time for low independent effective mass and temperatures and scattering time f requency for high t emperatures. The behavior of both the effective mass and the conductivity is in qualitative agreement with experimental r esul t s .

On the anharmonic, multi-phonon, Debye-Waller contributions to the phonon-limited resistivity of metals : applications to Na and K

The anharmonic, multi-phonon (MP), and Oebye-Waller factor (OW) contributions to the phonon limited resistivity (;0) of metals derived by Shukla and Muller (1979) by the doubletime temperature dependent Green function method have been numerically evaluated for Na and K in the high temperature limit. The anharmonic contributions arise from the cubic and quartic shift of phonons (CS, QS), and phonon width (W) and the interference term (1). The QS, MP and OW contributions to I' are also derived by the matrix element method and the results are in agreement with those of Shukla and Muller (1979). In the high temperature limit, the contributions to;O from each of the above mentioned terms are of the type BT2 For numerical calculations suitable expressions are derived for the anharmonic contributions to ~ in terms of the third and fourth rank tensors obtained by the Ewald procedure. The numerical calculation of the contributions to;O from the OW, MP term and the QS have been done exactly and from the CS, Wand I terms only approximately in the partial and total Einstein approximations (PEA, TEA), using a first principle approach (Shukla and Taylor (1976)). The results obtained indicate that there is a strong pairwise cancellation between the: OW and MP terms, the QS and CS and the Wand I terms. The sum total of these contributions to;O for Na and K amounts to 4 to 11% and 2 to 7%, respectively, in the PEA while in the TEA they amount to 3 to 7% and 1 to 4%, respectively, in the temperature range.

Comment le groupe informel agit sur la relation entre le leadership transformationnel et le climat organisationnel

Ce mémoire a pour objet l’influence du groupe informel sur la relation entre le leadership transformationnel et le climat organisationnel. Dans un premier temps, nous examinerons la théorie sur le leadership transformationnel. Le lien entre ce style de leadership et le climat organisationnel est à l’étude, car il semble que ces deux variables sont liés aux perceptions qu’ont les employés de leur vécu en organisation. Les résultats de notre étude confirment effectivement cette première hypothèse. L’objectif central de notre recherche consiste à voir comment le groupe informel agit sur la relation entre le leadership transformationnel et le climat organisationnel. D’abord, nous nous intéresserons sur l’appartenance à un groupe informel comme variable modératrice de la relation entre nos deux autres variables. Par contre, les résultats des analyses présentés n’ont pas réussi à confirmer ce modèle. Ensuite, nous étudierons le groupe informel en tant que variable médiatrice, où le chemin entre le leadership transformationnel et le climat organisationnel serait plutôt indirecte. Nos analyses démontrent que nos trois variables sont corrélées entre-elles, mais qu’en contrôlant pour l’effet du leadership transformationnel, le groupe informel n’est plus lié au climat organisationnel. Nous sommes amenés à proposer que le leadership transformationnel contribue à un contexte organisationnel qui satisfait aux besoins des employés, de sorte que le soutien social fourni par le groupe informel ne comble pas un manque. Ceci viendrait nuancer la dynamique à l’intérieur d’un groupe informel par rapport à ce qu’il apporte à ses membres.

Bootstrapping high frequency data

Nous développons dans cette thèse, des méthodes de bootstrap pour les données financières de hautes fréquences. Les deux premiers essais focalisent sur les méthodes de bootstrap appliquées à l’approche de "pré-moyennement" et robustes à la présence d’erreurs de microstructure. Le "pré-moyennement" permet de réduire l’influence de l’effet de microstructure avant d’appliquer la volatilité réalisée. En se basant sur cette ap- proche d’estimation de la volatilité intégrée en présence d’erreurs de microstructure, nous développons plusieurs méthodes de bootstrap qui préservent la structure de dépendance et l’hétérogénéité dans la moyenne des données originelles. Le troisième essai développe une méthode de bootstrap sous l’hypothèse de Gaussianité locale des données financières de hautes fréquences. Le premier chapitre est intitulé: "Bootstrap inference for pre-averaged realized volatility based on non-overlapping returns". Nous proposons dans ce chapitre, des méthodes de bootstrap robustes à la présence d’erreurs de microstructure. Particulièrement nous nous sommes focalisés sur la volatilité réalisée utilisant des rendements "pré-moyennés" proposés par Podolskij et Vetter (2009), où les rendements "pré-moyennés" sont construits sur des blocs de rendements à hautes fréquences consécutifs qui ne se chevauchent pas. Le "pré-moyennement" permet de réduire l’influence de l’effet de microstructure avant d’appliquer la volatilité réalisée. Le non-chevauchement des blocs fait que les rendements "pré-moyennés" sont asymptotiquement indépendants, mais possiblement hétéroscédastiques. Ce qui motive l’application du wild bootstrap dans ce contexte. Nous montrons la validité théorique du bootstrap pour construire des intervalles de type percentile et percentile-t. Les simulations Monte Carlo montrent que le bootstrap peut améliorer les propriétés en échantillon fini de l’estimateur de la volatilité intégrée par rapport aux résultats asymptotiques, pourvu que le choix de la variable externe soit fait de façon appropriée. Nous illustrons ces méthodes en utilisant des données financières réelles. Le deuxième chapitre est intitulé : "Bootstrapping pre-averaged realized volatility under market microstructure noise". Nous développons dans ce chapitre une méthode de bootstrap par bloc basée sur l’approche "pré-moyennement" de Jacod et al. (2009), où les rendements "pré-moyennés" sont construits sur des blocs de rendements à haute fréquences consécutifs qui se chevauchent. Le chevauchement des blocs induit une forte dépendance dans la structure des rendements "pré-moyennés". En effet les rendements "pré-moyennés" sont m-dépendant avec m qui croît à une vitesse plus faible que la taille d’échantillon n. Ceci motive l’application d’un bootstrap par bloc spécifique. Nous montrons que le bloc bootstrap suggéré par Bühlmann et Künsch (1995) n’est valide que lorsque la volatilité est constante. Ceci est dû à l’hétérogénéité dans la moyenne des rendements "pré-moyennés" au carré lorsque la volatilité est stochastique. Nous proposons donc une nouvelle procédure de bootstrap qui combine le wild bootstrap et le bootstrap par bloc, de telle sorte que la dépendance sérielle des rendements "pré-moyennés" est préservée à l’intérieur des blocs et la condition d’homogénéité nécessaire pour la validité du bootstrap est respectée. Sous des conditions de taille de bloc, nous montrons que cette méthode est convergente. Les simulations Monte Carlo montrent que le bootstrap améliore les propriétés en échantillon fini de l’estimateur de la volatilité intégrée par rapport aux résultats asymptotiques. Nous illustrons cette méthode en utilisant des données financières réelles. Le troisième chapitre est intitulé: "Bootstrapping realized covolatility measures under local Gaussianity assumption". Dans ce chapitre nous montrons, comment et dans quelle mesure on peut approximer les distributions des estimateurs de mesures de co-volatilité sous l’hypothèse de Gaussianité locale des rendements. En particulier nous proposons une nouvelle méthode de bootstrap sous ces hypothèses. Nous nous sommes focalisés sur la volatilité réalisée et sur le beta réalisé. Nous montrons que la nouvelle méthode de bootstrap appliquée au beta réalisé était capable de répliquer les cummulants au deuxième ordre, tandis qu’il procurait une amélioration au troisième degré lorsqu’elle est appliquée à la volatilité réalisée. Ces résultats améliorent donc les résultats existants dans cette littérature, notamment ceux de Gonçalves et Meddahi (2009) et de Dovonon, Gonçalves et Meddahi (2013). Les simulations Monte Carlo montrent que le bootstrap améliore les propriétés en échantillon fini de l’estimateur de la volatilité intégrée par rapport aux résultats asymptotiques et les résultats de bootstrap existants. Nous illustrons cette méthode en utilisant des données financières réelles.