The segmentation of medical images is a difficult task due to the inhomogeneous intensity variations that occurs during digital image acquisition, the complicated shape of the object, and the medical expert’s lack of semantic knowledge. Automated segmentation algorithms work well for some medical images, but no algorithm has been general enough to work for all medical images. In practice, most of the time the segmentation results are corrected by the experts before the actual use. In this work, we are motivated to determine how to make use of manually segmented data in automatic segmentation. The key idea is to transfer the ground truth segmentation from the database of train images to a given test image. The ground truth segmentation of MR images is done by experts. The process includes a hierarchical image decomposition approach that performs the shape matching of test images at several levels, starting with the image as a whole (i.e. level 0) and then going through a pyramid decomposition (i.e. level 1, level 2, etc.) with the database of the train images and the given test image. The goal of pyramid decomposition is to find the section of the training image that best matches a section of the test image of a different level. After that, a re-composition approach is taken to place the best matched sections of the training image to the original test image space. Finally, the ground truth segmentation is transferred from the best training images to their corresponding location in the test image. We have tested our method on a hip joint MR image database and the experiment shows successful results on level 0, level 1 and level 2 re-compositions. Results improve with deeper level decompositions, which supports our hypotheses.

Flowers, the sole natural source of pollen and nectar for bees, present many similar features, in colour, shape, size and scent, which facilitate pollinator attraction. This similarity among stimuli requires perception of commonality but also a capacity for differentiation between similar but different stimuli. While many flowers of a similar type will elicit approach and foraging, failure to access resources on any individual flower in an array (e.g. due to depletion) will not necessarily generalize and deter further foraging. Such conditions demand that bees respond to both the similarity and differences among stimuli which may share many common features but differ individually in available resources. Two questions are raised by this challenge and will herein be addressed: how do bees perceive and respond to ‘similar but different’ stimuli? And, how do bees use such cues to find rewarding flowers? Picture-object correspondence has not been previously specifically studied in invertebrates. The correspondence between picture-cue and object stimuli may offer a unique opportunity to trigger memory for corresponding targets while still retaining an important distinction between unrewarding cue and rewarding targets. Perception of pictures is not always perceived by animals as either the same as or equivalent to the intended subject. According to Fagot et al. (2000) the perceived relationship may result in confusion, independence or equivalence and is dependent upon experience. The objectives of this thesis are twofold: first, determine how bumblebees (Bombus impatiens) perceive the relationship between objects and corresponding pictures and secondly, to determine whether or not bees may be able to attend to and use pictures as cues while foraging. The correspondence of picture and object by bees was evaluated with four experiments of preference: (1) learned differentiation; spontaneous association to (2) colour, and (3) achromatic, impoverished images; and (4) learned picture cue use. Firstly, results show that bees do not confuse an object with a corresponding picture but nevertheless do perceive a relationship between them if colour cues are retained. Altered, achromatic images were not consistently treated as corresponding to coloured objects. Secondly, bees can learn to use a picture cue in a delayed matching foraging task. Results further suggest a role of three contributing factors in bumblebee picture cue use: (i) conditions of high inconsistency as to which target will be rewarding; (ii) stable target locations; and (iii) individual foraging experience. It appears that bumblebees can learn to use cues, in a delayed matching task, when the location of the corresponding target is known and stable, the individual bee has acquired some experience in successful foraging, and reward is otherwise unpredictable without the use of the cue. Bees may disregard secondary cues as noise under conditions of high target predictability whereby floral constancy or target perseveration may be most efficient, but attend to and learn such cues as signals if target reward is highly unpredictable. The conditions for this sensitivity may coincide with naturally occurring floral cycles.

Dense stereo matching is one of the most extensively investigated topics in computer vision, since it plays an important role in many applications such as 3D scene reconstruction. In this thesis, a novel dense stereo matching method is proposed based on edge-aware truncated minimum spanning tree (T-MST). Instead of employing non-local cost aggregation on traditional MST which is only generated from color differences of neighbouring pixels, a new tree structure, "Edge-Aware T-MST", is proposed to aggregate the cost according to the image texture. Specifically, cost aggregations are strongly enforced in large planar textureless regions due to the truncated edge weights. Meanwhile, the "edge fatten" effect is suppressed by employing a novel hybrid edge-prior which combines edge-prior and superpixel-prior to locate the potential disparity edges. Then a widely used Winner-Takes-All (WTA) strategy is performed to establish initial disparity map. An adaptive non-local refinement is also performed based on the stability of initial disparity estimation. Given the stereo images from Middlebury benchmark, we estimate the disparity maps by using our proposed method and other five state-of-the-art tree-based non-local matching methods. The experimental results show that the proposed method successfully produced reliable disparity values within large planar textureless regions and around object disparity boundaries. Performance comparisons demonstrate that our proposed non-local stereo matching method based on edge-aware T-MST outperforms current non-local tree-based state-of-the-art stereo matching methods in most cases, especially in large textureless planar regions and around disparity bounaries.

Hydrogen is a potential future energy storage medium to supplement a variety of renewable energy sources. It can be regarded as an environmentally-friendly fuel, especially when it is extracted from water using electricity obtained from solar panels or wind turbines. The focus in this thesis is on solar energy, and the theoretical background (i.e., PSCAD computer simulation) and experimental work related to a water-splitting, hydrogen-production system are presented. The hydrogen production system was powered by a photovoltaic (PV) array using a proton exchange membrane (PEM) electrolyser. The PV array and PEM electrolyser display an inherently non-linear current-voltage relationship that requires optimal matching of maximum operating power. Optimal matching between the PV system and the electrolyser is essential to maximise the transfer of electrical energy and the rate of hydrogen production. A DC/DC converter is used for power matching by shifting the PEM electrolyser I-V curve as closely as possible toward the maximum power the PV can deliver. By taking advantage of the I-V characteristics of the electrolyser (i.e., the DC/DC converter output voltage is essentially constant whereas the current increases dramatically), we demonstrated experimentally and in simulations that the hydrogen production of the PV-electrolyser system can be optimised by adjusting the duty cycle generated by the pulse-width modulation (PWM) circuit. The strategy used was to fix the duty cycle at the ratio of the PV maximum power voltage to the electrolyser operating voltage. A stand-alone PV energy system, using hydrogen as the storage medium, was designed. The system would be suitable for providing power for a family's house located in a remote area in the Libyan Sahara.

Thesis advisor: M. Utku Ünver / Thesis advisor: M. Bumin Yenmez / This dissertation consists of two essays in market design. In the first chapter, we study affirmative action policies in college admissions and hiring. A college or firm makes admissions or hiring decisions in which each candidate is characterized by priority ranking and type, which may depend on race, gender, or socioeconomic status. The admissions or hiring committee faces a trade-off between meritocracy and diversity: while a merit-first choice rule may admit candidates of the same type, a diversity-first choice rule may be unfair due to priority violations. To formalize this trade-off, we introduce a measure of meritocracy and a measure of diversity for choice rules. Then, we investigate how to resolve the tension between them. A choice rule that uses both reserves and quotas can be viewed as a compromise and is a generalization of the two extreme rules. The first result is comparative statics for this class of choice rules: we show that as parameters change and the choice rule becomes more meritorious, it also becomes less diverse. The second result is a characterization of the choice rule, which may help admissions or hiring committees to decide their policies. In the second chapter, we introduce a method to measure manipulability of a matching mechanism and use theory and simulation to study constrained mechanisms in school choice. First, we show that the implications from existing measures are strongly dependent on the full preference domain assumption. Our measure is more robust. The implications from existing measures can be carried over as well: while the recent school admissions reforms did not fully eliminate incentives to manipulate, they discouraged manipulation. Second, we use simulations for quantitative analysis. Our results support the recent school admissions reforms quantitatively, as well as qualitatively: they largely eliminated the incentives to manipulate. In addition, while the qualitative implications from theory are parallel to existing measures, the quantitative implications from simulations confirm a significant difference. / Thesis (PhD) — Boston College, 2021. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Economics.

Due to the amount of flow simulation and measurement data, automatic detection, classification and visualization of features is necessary for an inspection. Therefore, many automated feature detection methods have been developed in recent years. However, one feature class is visualized afterwards in most cases, and many algorithms have problems in the presence of noise or superposition effects. In contrast, image processing and computer vision have robust methods for feature extraction and computation of derivatives of scalar fields. Furthermore, interpolation and other filter can be analyzed in detail. An application of these methods to vector fields would provide a solid theoretical basis for feature extraction. The authors suggest Clifford algebra as a mathematical framework for this task. Clifford algebra provides a unified notation for scalars and vectors as well as a multiplication of all basis elements. The Clifford product of two vectors provides the complete geometric information of the relative positions of these vectors. Integration of this product results in Clifford correlation and convolution which can be used for template matching on vector fields. Furthermore, for frequency analysis of vector fields and the behavior of vector-valued filters, a Clifford Fourier transform has been derived for 2 and 3 dimensions. Convolution and other theorems have been proved, and fast algorithms for the computation of the Clifford Fourier transform exist. Therefore the computation of Clifford convolution can be accelerated by computing it in Clifford Fourier domain. Clifford convolution and Fourier transform can be used for a thorough analysis and subsequent visualization of vector fields

In dieser Arbeit wird Match-R-CNN unter dem Gesichtspunkt des Product-Matchings analysiert und implementiert. Bei Match-R-CNN handelt es sich um ein Framework, welches zur Analyse von Bekleidungsbildern eingesetzt werden kann. Es wurde bei Ge et al. [GZW+19] eingeführt. Product-Matching ist die Aufgabe zwei identische Produkte zu identifizieren. Methoden der Bildverabeitung und maschinellen Lernens werden erläutert. Des Weiteren wird der aktuelle Forschungsstand in verwandten Gebieten erörtert. Es war möglich den Aufbau von Match-R-CNN zu analysieren. Hierfür wurden Ge et al. [GZW+19] und Diskussionen im dazugehörigen Github-Repository [git19] herangezogen. Um die Implementierung abschließend zu bewerten, ist weitere Arbeit notwendig.:1 Einleitung 2 Grundlagen 2.1 Bildverarbeitung 2.1.1 Kantenerkennung 2.1.2 Bildfaltung 2.1.3 Probleme bei der Umsetzung 2.2 Convolutional Neural Networks 2.2.1 Probleme bei konventionellen künstlichen neuronalen Netzen 2.2.2 Besonderheiten bei CNNs 2.2.3 Aufbau und Hyperparameter 2.2.4 Training von CNNs 2.2.5 Aktuelle Erkenntnisse 2.3 Ähnlichkeit auf Bildern 3 Verwandte Arbeiten 3.1 Clothing Retrieval und Detection 3.2 Product-Matching 3.3 Deep Similarity 4 Methodik und Umsetzung 4.1 Datensatz 4.2 Datenaufbereitung 4.3 Netzwerkarchitektur 4.3.1 Feature-Network 4.3.2 Matching-Network 4.4 Strategie zur Erzeugung der Trainingspaare 4.5 Matching-Network Training 4.6 Experimente und Zwischenergebnisse 4.7 Ergebnisse 5 Fazit 6 Ausblick Literaturverzeichnis Abbildungsverzeichnis

As biomedical research and advances in biotechnology generate expansive datasets, the need to process this data into information has grown simultaneously. Specifically, recognizing and extracting these “key” phrases comprising the named entities from this information databank promises a plethora of applications for scientists. The ability to construct interaction maps,identify proteins as drug targets are two important applications. Since we have the choice of defining what is “useful”, we can potentially utilize text mining for our purpose. In a novel attempt to beat the challenge, we have put information theory and text compression through this task. Prediction by partial matching is an adaptive text encoding scheme that blends together a set of finite context Markov models to predict the probability of the next token in a given symbol stream. We observe, named entities such as gene names, protein names, gene functions, protein-protein interactions – all follow symbol statistics uniquely different from normal scientific text. By using well defined training sets that allow us to selectively differentiate between named entities and the rest of the symbols; we were able to extract them with a good accuracy. We have implemented our tests, using the Text Mining Toolkit, on identification of gene functions and protein-protein interactions with f-scores (based on precision & recall) of 0.9737 and 0.6865 respectively. With our results, we foresee the application of such an approach in automated information retrieval in the realm of biology.

Ontologies describe the semantics of data and provide a uniform framework of understanding between different parties. The main common reference to an ontology definition describes them as knowledge bodies, which bring a formal representation of a shared conceptualization of a domain - the objects, concepts and other entities that are assumed to exist in a certain area of interest together with the relationships holding among them. However, in open and evolving systems with decentralized nature (as, for example, the Semantic Web), it is unlikely for different parties to adopt the same ontology. The problem of ontology matching evolves from the need to align ontologies, which cover the same or similar domains of knowledge. The task is to reducing ontology heterogeneity, which can occur in different forms, not in isolation from one another. Syntactically heterogeneous ontologies are expressed in different formal languages. Terminological heterogeneity stands for variations in names when referring to the same entities and concepts. Conceptual heterogeneity refers to differences in coverage, granularity or scope when modeling the same domain of interest. Finally, prgamatic heterogeneity is about mismatches in how entities are interpreted by people in a given context. The work presented in this thesis is a contribution to the problem of reducing the terminological and conceptual heterogeneity of hierarchical ontologies (defined as ontologies, which contain a hierarchical body), populated with text documents. We make use of both intensional (structural) and extensional (instance-based) aspects of the input ontologies and combine them in order to establish correspondences between their elements. In addition, the proposed procedures yield assertions on the granularity and the extensional richness of one ontology compared to another, which is helpful at assisting a process of ontology merging. Although we put an emphasis on the application of instance-based techniques, we show that combining them with intensional approaches leads to more efficient (both conceptually and computationally) similarity judgments. The thesis is oriented towards both researchers and practitioners in the domain of ontology matching and knowledge sharing. The proposed solutions can be applied successfully to the problem of matching web-directories and facilitating the exchange of knowledge on the web-scale.

Covariate quality has been primarily theory driven in propensity score matching with a general adversity to the interpretation of group prediction. However, effect sizes are well supported in the literature and may help to inform the method. Specifically, I index can be used as a measure of effect size in logistic regression to evaluate group prediction. As such, simulation was used to create 35 conditions of I, initial bias and sample size to examine statistical differences in (a) post-matching bias reduction and (b) treatment effect sensitivity. The results of this study suggest these conditions do not explain statistical differences in percent bias reduction of treatment likelihood after matching. However, I and sample size do explain statistical differences in treatment effect sensitivity. Treatment effect sensitivity was lower when sample sizes and I increased. However, this relationship was mitigated within smaller sample sizes as I increased above I = .50.