No description available.

This thesis investigates two well defined problems in image segmentation, viz. interactive and semantic image segmentation. Interactive segmentation involves power assisting a user in cutting out objects from an image, whereas semantic segmentation involves partitioning pixels in an image into object categories. We investigate various models and energy formulations for both these problems in this thesis. In order to improve the performance of interactive systems, low level texture features are introduced as a replacement for the more commonly used RGB features. To quantify the improvement obtained by using these texture features, two annotated datasets of images are introduced (one consisting of natural images, and the other consisting of camouflaged objects). A significant improvement in performance is observed when using texture features for the case of monochrome images and images containing camouflaged objects. We also explore adding mid-level cues such as shape constraints into interactive segmentation by introducing the idea of geodesic star convexity, which extends the existing notion of a star convexity prior in two important ways: (i) It allows for multiple star centres as opposed to single stars in the original prior and (ii) It generalises the shape constraint by allowing for Geodesic paths as opposed to Euclidean rays. Global minima of our energy function can be obtained subject to these new constraints. We also introduce Geodesic Forests, which exploit the structure of shortest paths in implementing the extended constraints. These extensions to star convexity allow us to use such constraints in a practical segmentation system. This system is evaluated by means of a “robot user” to measure the amount of interaction required in a precise way, and it is shown that having shape constraints reduces user effort significantly compared to existing interactive systems. We also introduce a new and harder dataset which augments the existing GrabCut dataset with more realistic images and ground truth taken from the PASCAL VOC segmentation challenge. In the latter part of the thesis, we bring in object category level information in order to make the interactive segmentation tasks easier, and move towards fully automated semantic segmentation. An algorithm to automatically segment humans from cluttered images given their bounding boxes is presented. A top down segmentation of the human is obtained using classifiers trained to predict segmentation masks from local HOG descriptors. These masks are then combined with bottom up image information in a local GrabCut like procedure. This algorithm is later completely automated to segment humans without requiring a bounding box, and is quantitatively compared with other semantic segmentation methods. We also introduce a novel way to acquire large quantities of segmented training data relatively effortlessly using the Kinect. In the final part of this work, we explore various semantic segmentation methods based on learning using bottom up super-pixelisations. Different methods of combining multiple super-pixelisations are discussed and quantitatively evaluated on two segmentation datasets. We observe that simple combinations of independently trained classifiers on single super-pixelisations perform almost as good as complex methods based on jointly learning across multiple super-pixelisations. We also explore CRF based formulations for semantic segmentation, and introduce novel visual words based object boundary description in the energy formulation. The object appearance and boundary parameters are trained jointly using structured output learning methods, and the benefit of adding pairwise terms is quantified on two different datasets.

O problema da Satisfazibilidade Probabilística, PSAT, apesar da sua flexibilidade, torna exponencialmente complexa a modelagem de variáveis estatisticamente independentes. Esta tese busca desenvolver algoritmos e propostas de relaxamento para permitir o tratamento eficiente de independência parcial pelo PSAT. Apresentamos uma aplicação do PSAT ao problema da etiquetagem morfossintática que serve tanto de motivação como de demonstração dos conceitos apresentados. / The Probabilistic Satisfiability Problem, PSAT, despite its flexibility, makes it exponentially complicated to model statistically independent variables. This thesis develops algorithms and relaxation proposals that allow an efficient treatment of partial independence with PSAT. We also present an application of PSAT on the Part-of-speech tagging problem to serve both as motivation and showcase of the presented concepts.

[pt] Em 1988, Delzant classificou as variedades compactas tóricas simpléticas por meio da imagem associada da aplicação momento. Como estabelecido pelo Teorema de Convexidade [Atiyah, Guillemin-Sternberg, 1983], a imagem pela aplicação momento de uma variedade compacta tórica simplética é um polítopo convexo. A construção de Delzant proporciona uma receita para formar, dado um polítopo de Delzant, uma variedade compacta tórica simplética. Nesta dissertação revisamos essa construção e estudamos alguns exemplos. / [en] In 1988, Delzant proved a classification Theorem of compact toric symplectic manifolds by means of their moment image. By the convexity Theorem [Atiyah, Guillemin-Sternberg, 1983] the moment image of a compact toric symplectic manifold is a convex polytope. Delzant s construction gives a recipe to construct, given a Delzant polytope, the corresponding compact toric symplectic manifold. This thesis describes this construction and studies in detail some examples.

Cette thèse porte sur l’étude des propriétés structurelles de graphes dont la compréhension permet de concevoir des algorithmes efficaces pour résoudre des problèmes d’optimisation. Nous nous intéressons plus particulièrement aux méthodes de décomposition des graphes, aux jeux de poursuites et à la notion de convexité. Le jeu de Processus a été défini comme un modèle de la reconfiguration de routage. Souvent, ces jeux où une équipe de chercheurs doit effacer un graphe non orienté sont reliés aux décompositions de graphes. Dans les digraphes, nous montrons que le jeu de Processus est monotone et nous définissons une nouvelle décomposition de graphes que lui est équivalente. Ensuite, nous étudions d’autres décompositions de graphes. Nous proposons un algorithme FPT-unifiée pour calculer plusieurs paramètres de largeur de graphes. En particulier, ceci est le premier FPT-algorithme pour la largeur arborescente q-branché et spéciale d’un graphe. Nous étudions ensuite un autre jeu qui modélise les problèmes de pré-chargement. Nous introduisons la variante en ligne du jeu de surveillance. Nous étudions l’écart entre le jeu de surveillance classique et ses versions connecté et en ligne, en fournissant de nouvelles bornes. Nous définissons ensuite un cadre général pour l’étude des jeux poursuite-évasion. Cette méthode nous permet de donner les premiers résultats d’approximation pour certains de ces jeux. Finalement, nous étudions un autre paramètre lié à la convexité des graphes et à la propagation d’infection dans les réseaux, le nombre enveloppe. Nous fournissons plusieurs résultats de complexité en fonction des structures des graphes et en utilisant des décompositions de graphes. / This thesis focuses on the study of structural properties of graphs whose understanding enables the design of efficient algorithms for solving optimization problems. We are particularly interested in methods of decomposition, pursuit-evasion games and the notion of convexity. The Process game has been defined as a model for the routing reconfiguration problem in WDM networks. Often, such games where a team of searchers have to clear an undirected graph are closely related to graph decompositions. In digraphs, we show that the Process game is monotone and we define a new equivalent digraph decomposition. Then, we further investigate graph decompositions. We propose a unified FPT-algorithm to compute several graph width parameters. This algorithm turns to be the first FPT-algorithm for the special and the q-branched tree-width of a graph. We then study another pursuit-evasion game which models prefetching problems. We introduce the more realistic online variant of the Surveillance game. We investigate the gap between the classical Surveillance Game and its connected and online versions by providing new bounds. We then define a general framework for studying pursuit-evasion games, based on linear programming techniques. This method allows us to give first approximation results for some of these games. Finally, we study another parameter related to graph convexity and to the spreading of infection in networks, namely the hull number. We provide several complexity results depending on the graph structures making use of graph decompositions. Some of these results answer open questions of the literature.

本研究探討國內債券市場如何決定債券殖利率之風險貼水及交易成本貼水，其中風險貼水可分為時間及信用風險貼水兩部份，屬於前者之重要變數有存續期間(Duration)與凸性(Convexity)，屬於後者則有信用評等(Credit Ranking)與銀行擔保效果，而交易成本分析是在控制風險貼水因素後，看稅賦效果是否會影響投資者之必要報酬率。此外，本研究亦對殖利率曲線作分析，討論長短期資金市場是否存在明顯互動關係。首先對存續期間及凸性之特性作研究，檢測在既有到期期間變數下，加入存續期間及凸性對於債券殖利率邊際解釋能力的影響，冀描述國內投資者之訂價行為。其次就債券信用評等與銀行擔保效果作分析，公司債可能有不同信用等級，面對不等級公司債券，投資人或會要求不同程度之違約風險貼水，本文以實證對此作探討。又公司債因擔保與否，區分為擔保公司債及無擔保公司債，因此本研究關心的第二組變數是發行人為政府抑民間機構暨公司債之信用評等與銀行擔保效果。本研究亦對債券交易成本作探討，就證券交易稅而言，政府公債免徵交易稅而公司債券買賣須課徵千分之一的交易稅，因此我們想要知道存在公債與公司債之間，因稅賦差異造成交易成本不同，是否會影響到投資者的意願，故第三項變數為以證交稅為主之交易成本。最後對市場資金供需情形作分析，同時探討長短期資金市場是否存在互動關係，故第四項變數為全面資金供需情況。本研究藉檢測總體經濟資金供需變數對債券殖利率之影響，同時也可檢驗國內長短期資金市場之區隔程度。由於國內債市尚淺，仍舊有很大的發展空間，因此在可預期的未來，國內債券市場勢必會受到應有的重視。但由於早期國內債市之不發達及不受重視，使得有關債券資料的保存，特別是公司債券部份十分缺乏，連帶相關文獻亦寥寥可數，故本文以國內債券市場資料做實證研究、分析，冀望能有邊際之貢獻。

Στη διατριβή αυτή μελετάται η μονοτονία και η κυρτότητα των ριζών ορισμένων οικογενειών associated ορθογωνίων q-πολυωνύμων που εμφανίζονται στο q-ανάλογο του σχήματος Askey. Για τη μελέτη της μονοτονίας και της κυρτότητας των ριζών χρησιμοποιείται μια συναρτησιακή αναλυτική μέθοδος η οποία βασίζεται στην αναδρομική σχέση τριών όρων που ικανοποιεί οποιαδήποτε οικογένεια ορθογωνίων πολυωνύμων. Επίσης για τον υπολογισμό των αθροισμάτων Newton των ριζών χρησιμοποιείται η συναρτησιακή αναλυτική μλεθοδος που παρουσιάστηκε από τους Υφαντή, Κοκολογιαννάκη και Σιαφαρίκα για τον υπολογισμό των αθροισμάτων Newton των ριζών των scaled corecursive associated ορθογωνίων πολυωνύμων. Επειδή τα ορθογώνια q-πολυώνυμα είναι q-ανάλογα κλασικών ορθογωνίων πολυωνύμων παίρνοντας το όριο q-1 προκύπτουν αντίστοιχα αποτελέσματα για τις ρίζες των κλασσικών ορθογωνίων πολυωνύμων. Τα αποτελέσματα αυτά γενικεύουν ενοποιούν και βελτιώνουν προηγούμενα αποτελέσματα. / In this thesis, we study the monotonicity properties and the convexity of the zeros of some families of associated orthogonal q-polynomials. Also, we calculate the Newton sum rules of these zeros. For the study of the monotonicity of the zeros, we use a functional analytic method based on the three terms recurrence relations satisfied by the associated orthogonal q-polynomials under consideration.

Der menschliche inferiore Parietallappen (IPC) gehört zum Assoziationskortex und spielt eine wichtige Rolle bei der Integration von somatosensorischen (taktilen), visuellen und akustischen Reizen. Bisher gibt es keine eindeutigen Informationen über den strukturellen Aufbau dieser Hirnregion. Parzellierungen anhand der Zytoarchitektur reichen von zwei (Brodmann 1909) bis sieben Subareale (Caspers et al. 2006). Homologien zwischen dem IPC des Menschen und Makaken-Affen sind weitestgehend unbekannt. In der vorliegenden Arbeit wurden der Aufbau und die Konnektivitäten des menschlichen IPC genauer untersucht. Dazu führte man eine konnektivitätsbasierte Parzellierung des IPC an 20 Probanden durch. Als Methode kam Diffusions-Tensor-Imaging (DTI) kombiniert mit probabilistischer Traktogra-phie zum Einsatz. Der IPC konnte anhand der Konnektivitäten in drei Subareale (IPCa, IPCm, IPCp) parzelliert werden. Diese besitzen in beiden Hemisphären eine ähnliche Größe und eine rostro-kaudale Anordnung. Die Parzellierung ist vergleichbar mit der des Makaken-IPC, bei dem ebenfalls eine Unterteilung in drei Areale (PF, PFG, PG) und eine rostro-kaudale Anordnung nachgewiesen werden konnte. Jedes Subareal des menschlichen IPC besitzt ein individuelles Konnektivitätsmuster. Beim Menschen als auch beim Makaken gibt es starke Verbindungen zum lateralen prämotorischen Kortex und zum superioren Parietallappen. Diese Gemeinsamkeiten lassen darauf schließen, dass strukturelle Eigenschaften im Laufe der Evolution erhalten geblieben sind. Allerdings sind beim Menschen auch Neuentwicklungen nachweisbar. Dazu gehören die deutlich hervortretenden Verbindungen zum Temporallappen. Möglicherweise haben sich diese erst während der Evolution entwickelt und sind beim Menschen als Teil des perisylvischen Sprachnetzwerkes an der Sprachbildung beteiligt. / The human inferior parietal cortex convexity (IPCC) is an important association area, which integrates auditory, visual and somatosensory information. However, the structural organization of the IPCC is a controversial issue. For example, cytoarchitectonic parcellations reported in the literature range from two to seven areas. Moreover, anatomical descriptions of the human IPCC are often based on experiments in the macaque monkey. In this study we used diffusion-weighted magnetic resonance imaging (dMRI) combined with probabilistic tractography to quantify the connectivity of the human IPCC, and used this information to parcellate this cortex area. This provides a new structural map of the human IPCC, comprising three sub-areas (IPCa, IPCm, IPCp) of comparable size, in a rostro-caudal arrangement in the left and right hemisphere. Each sub-area is characterized by a connectivity fingerprint and the parcellation is similar to the subdivision reported for the macaque IPCC (rostro-caudal areas areas PF, PFG, and PG). However, the present study also reliably demonstrates new structural features in the connectivity pattern of the human IPCC, which are not known to exist in the macaque. This study quantifies inter-subject variability by providing a population representation of the sub-area arrangement, and demonstrates substantial lateralization of the connectivity patterns of IPCC.

The thesis consists of four papers. The first two essays deal with entry externalities, the third studies the Law of One Price (LOP), while the last essay examines average profits for a monopolist under uncertainty. In the first essay, entry externalities in the form of information and positive payoff externalities are studied. When a firm enters a market, it often imposes externalities on existing firms and/or future potential entrants. If products are substitutes, these externalities are typically negative; if products are complements, the externalities are typically positive. Externalities related to substitution or complementarities between products are called payoff externalities, since entry by one firm has a direct effect on the other firms' payoff. Another type of externality arises when firms have private information about the profitability of entry. In this case, the entry decision of one firm potentially reveals that firm's private information. The focus of the paper is on the scope for intervention for an uninformed social planner, when firms privately know the profitability of entry and moreover, the firms have an option to delay their entry. The main result is that there is insufficient entry, since firms delay too much in equilibrium and further, the social planner can increase welfare by subsidizing early entry. Continuing on this theme, the second essay has the same focus, but instead takes the time of entry as fixed, while generalizing the analysis of payoff externalities also to the case of negative payoff externalities. The main contribution is the characterization of equilibria under both positive and negative payoff externalities and the implications for public policy. Here, the scope for intervention will, in contrast to the results in the first essay, be low, when entry is profitable for uninformed firms. In the third essay (joint with Richard Friberg), deviations from the LOP are studied in the presence of transport costs, under the assumption that firms can endogenously choose to segment markets in order to prevent arbitrage by consumers. It is shown that the deviation from LOP can increase as transport costs fall between countries. The last essay (joint with Richard Friberg), studies the problem facing a monopolist when the cost of inputs is uncertain. The main result is that the monopolist can gain from this uncertainty, in the sense that average profits are increasing in the variability of costs. / Diss. Stockholm : Handelshögsk., 2001

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / In this dissertation we present a proof of a Bernard Malgrange theorem, which establishes a necessary and sufficient condition for the global solvability of a linear operator with constant coefficients. / Nessa disserta¸c ao apresentamos uma demonstra¸c ao do Teorema de Bernard Malgrange, que estabelece condi¸c ao necess´aria e suficiente para que um operador linear com coeficientes constantes seja globalmente resol´uvel.