We develop a new autoregressive conditional process to capture both the changes and the persistency of the intraday seasonal (U-shape) pattern of volatility in essay 1. Unlike other procedures, this approach allows for the intraday volatility pattern to change over time without the filtering process injecting a spurious pattern of noise into the filtered series. We show that prior deterministic filtering procedures are special cases of the autoregressive conditional filtering process presented here. Lagrange multiplier tests prove that the stochastic seasonal variance component is statistically significant. Specification tests using the correlogram and cross-spectral analyses prove the reliability of the autoregressive conditional filtering process. In essay 2 we develop a new methodology to decompose return variance in order to examine the informativeness embedded in the return series. The variance is decomposed into the information arrival component and the noise factor component. This decomposition methodology differs from previous studies in that both the informational variance and the noise variance are time-varying. Furthermore, the covariance of the informational component and the noisy component is no longer restricted to be zero. The resultant measure of price informativeness is defined as the informational variance divided by the total variance of the returns. The noisy rational expectations model predicts that uninformed traders react to price changes more than informed traders, since uninformed traders cannot distinguish between price changes caused by information arrivals and price changes caused by noise. This hypothesis is tested in essay 3 using intraday data with the intraday seasonal volatility component removed, as based on the procedure in the first essay. The resultant seasonally adjusted variance series is decomposed into components caused by unexpected information arrivals and by noise in order to examine informativeness.

This paper is concerned with approximating noisy samples by non-uniform rational B-spline curves with special emphasis on free knots. We show how to set up the problem such that nonlinear optimization methods can be applied efficiently. This involves the introduction of penalizing terms in order to avoid undesired knot positions. We report on our implementation of the nonlinear optimization and we show a way to implement the program in parallel. Parallel performance results are described. Our experiments show that our program has a linear speedup and an efficiency value close to unity. Runtime results on a parallel computer are displayed.

Cette thèse montre des contributions à l'analyse d'algorithmes pour l'optimisation de fonctions bruitées. Les taux de convergences (regret simple et regret cumulatif) sont analysés pour les algorithmes de recherche linéaire ainsi que pour les algorithmes de recherche aléatoires. Nous prouvons que les algorithmes basé sur la matrice hessienne peuvent atteindre le même résultat que certaines algorithmes optimaux, lorsque les paramètres sont bien choisis. De plus, nous analysons l'ordre de convergence des stratégies évolutionnistes pour des fonctions bruitées. Nous déduisons une convergence log-log. Nous prouvons aussi une borne basse pour le taux de convergence de stratégies évolutionnistes. Nous étendons le travail effectué sur les mécanismes de réévaluations en les appliquant au cas discret. Finalement, nous analysons la mesure de performance en elle-même et prouvons que l'utilisation d'une mauvaise mesure de performance peut mener à des résultats trompeurs lorsque différentes méthodes d'optimisation sont évaluées. / This thesis exposes contributions to the analysis of algorithms for noisy functions. It exposes convergence rates for linesearch algorithms as well as for random search algorithms. We prove in terms of Simple Regret and Cumulative Regret that a Hessian based algorithm can reach the same results as some optimal algorithms in the literature, when parameters are tuned correctly. On the other hand we analyse the convergence order of Evolution Strategies when solving noisy functions. We deduce log-log convergence. We also give a lower bound for the convergence rate of the Evolution Strategies. We extend the work on revaluation by applying it to a discrete settings. Finally we analyse the performance measure itself and prove that the use of an erroneus performance measure can lead to misleading results on the evaluation of different methods.

Clustering is an unsupervised learning technique widely used to group data into homogeneous clusters. For many real-world data containing categorical values, existing algorithms are often computationally costly in high dimensions, do not work well on noisy data with missing values, and rarely provide theoretical guarantees on clustering accuracy. In this thesis, we propose a general categorical data encoding method and a computationally efficient spectral based algorithm to cluster high-dimensional noisy categorical (nominal or ordinal) data. Under a statistical model for data on m attributes from n subjects in r clusters with missing probability epsilon, we show that our algorithm exactly recovers the true clusters with high probability when mn(1-epsilon) >= CMr² log³M, with M=max(n,m) and a fixed constant C. Moreover, we show that mn(1- epsilon)² >= r *delta/2 with 0< delta <1 is necessary for any algorithm to succeed with probability at least (1+delta)/2. In case, where m=n and r is fixed, for example, the sufficient condition matches with the necessary condition up to a polylog(n) factor, showing that our proposed algorithm is nearly optimal. We also show our algorithm outperforms several existing algorithms in both clustering accuracy and computational efficiency, both theoretically and numerically. In addition, we propose a spectral algorithm with standardization to cluster mixed data. This algorithm is computationally efficient and its clustering accuracy has been evaluated numerically on both real world data and synthetic data.

No description available.

Training a neural network requires a large amount of labeled data that has to be created by either human annotation or by very specifically created methods. Currently, there is a vast abundance of unlabeled data that is neglected sitting on servers, hard drives, websites, etc. These untapped data sources serve as the inspiration for this paper. The goal of this thesis is to explore and test various methods of semi-supervised learning (SSL) for convolutional neural networks (CNN). These methods will be analyzed and evaluated based on their accuracy on a test set of data. Since this particular neural network will be used to offer paths for an autonomous robot, it is important for the networks to be lightweight in scale. This paper will then take this assortment of smaller neural networks and run them through a variety of semi-supervised training methods. The first method is to have a teacher model that is trained on properly labeled data create labels for unlabeled data and add this to the training set for the next student model. From this base method, a few variations were tried in the hopes of getting a significant improvement. The first variation tested by this thesis is the effects of having this teacher and student cycle run more than one iteration. After that, the effects of using the confidence values that the models produced were explored by both including only data with confidence above a certain value and in a different test, relabeling data below a confidence threshold. The last variation this thesis explored was to have two teacher models concurrently and have the combination of those two models decide on the proper label for the unlabeled data. Through exploration and testing, these methods are evaluated in the results section as to which one produces the best results for SSL.

No description available.

No description available.

In this work we present a new tool for the convergence analysis of numerical optimization methods. It is based on the concepts of the Clarke derivative and set-valued mappings. Our goal is to apply this tool to minimization problems with non-smooth and noisy objective functions. After deriving a necessary condition for minimizers of such functions, we examine two unconstrained optimization routines. First, we prove new convergence theorems for Implicit Filtering and General Pattern Search. Then we show how these results can be used in practice, by executing some numerical computations. / Master of Science

En prenant particulièrement appui sur le cas de la cité de maisons préfabriquées du Merlan à Noisy-le-Sec, cette thèse élabore une critique du processus de patrimonialisation. Elle en met en évidence les enjeux sociaux, architecturaux et historiques. La nature singulière des petites constructions de la cité qui date de la Reconstruction, amène à interroger les termes mêmes sous lesquels elles sont aujourd'hui considérées : s'agit-il d'un monument à considérer comme un ensemble ou de monuments historiques divers ? D'un patrimoine ou de patrimoines ? Le quartier, par son histoire et ses spécificités, oblige à reconsidérer des termes dont la définition, pour commune qu'elle semble, ne va pas sans histoire ni problèmes. La thèse l'établit à un certain niveau de généralité en évoquant d'autres cas, plus ou moins proches de ceux de la cité. Le but est d'esquisser des propositions qui tiennent compte du caractère habité, hier comme aujourd'hui, des cas considérés. Il s'agit d'abord de proposer des concepts aptes à déplacer le champ des perspectives et polémiques classiques (préserver, conserver, restaurer) : ceux de «traduction» et de «mises en récits». Il s'agit ensuite d'éprouver, de réaliser ou de mettre en œuvre ces concepts en mettant en jeu les techniques du «web-documentaire». L'objectif est d'exposer une forme alternative de recherche, théorique et pratique, susceptible de permettre à un public élargi de s'interroger sur la mémoire des lieux. Le travail concerne ainsi le design qu'il envisage lui-même comme élément méthodologique de recherche. / Focusing on the case of the prefabricated houses in the cité of Merlan in Noisy-le-Sec, this PhD analyses the process of patriomonialisation or cultural heritage. It underlines its social, architectural and historical aspects. The singularity of the small buildings of the cité which date back to the Reconstruction period makes us wonder about the terms to be used to refer to it: is it a monument on its own or a set or various historical monuments? Should we talk about a heritage or heritages in that case? Because of its history and specificities, the neighbourhood forces us to reconsider the definitions of supposedly-familiar words. This thesis aims at developing these definitions to a certain level of generality also using other cases, quite similar to those of the cité. The purpose is to suggest projects that take into account the fact that these buildings were lived in in the past and are still lived in today. First, it proposes some concepts able to move the perspectives and the conventional controversy (preserve, conserve, restore): those of "translating" and "putting into stories". Then it experiments, realizes and implements these concepts on a practical level by using "web-documentary" techniques. The aim is to expose an alternative research form, both theoretical and practical, potentially enabling a wider audience to question the memory of places. This work involves design and considers it a methodological research element.