Learning from Geometry

Subspaces and manifolds are two powerful models for high dimensional signals. Subspaces model linear correlation and are a good fit to signals generated by physical systems, such as frontal images of human faces and multiple sources impinging at an antenna array. Manifolds model sources that are not linearly correlated, but where signals are determined by a small number of parameters. Examples are images of human faces under different poses or expressions, and handwritten digits with varying styles. However, there will always be some degree of model mismatch between the subspace or manifold model and the true statistics of the source. This dissertation exploits subspace and manifold models as prior information in various signal processing and machine learning tasks.

A near-low-rank Gaussian mixture model measures proximity to a union of linear or affine subspaces. This simple model can effectively capture the signal distribution when each class is near a subspace. This dissertation studies how the pairwise geometry between these subspaces affects classification performance. When model mismatch is vanishingly small, the probability of misclassification is determined by the product of the sines of the principal angles between subspaces. When the model mismatch is more significant, the probability of misclassification is determined by the sum of the squares of the sines of the principal angles. Reliability of classification is derived in terms of the distribution of signal energy across principal vectors. Larger principal angles lead to smaller classification error, motivating a linear transform that optimizes principal angles. This linear transformation, termed TRAIT, also preserves some specific features in each class, being complementary to a recently developed Low Rank Transform (LRT). Moreover, when the model mismatch is more significant, TRAIT shows superior performance compared to LRT.

The manifold model enforces a constraint on the freedom of data variation. Learning features that are robust to data variation is very important, especially when the size of the training set is small. A learning machine with large numbers of parameters, e.g., deep neural network, can well describe a very complicated data distribution. However, it is also more likely to be sensitive to small perturbations of the data, and to suffer from suffer from degraded performance when generalizing to unseen (test) data.

From the perspective of complexity of function classes, such a learning machine has a huge capacity (complexity), which tends to overfit. The manifold model provides us with a way of regularizing the learning machine, so as to reduce the generalization error, therefore mitigate overfiting. Two different overfiting-preventing approaches are proposed, one from the perspective of data variation, the other from capacity/complexity control. In the first approach, the learning machine is encouraged to make decisions that vary smoothly for data points in local neighborhoods on the manifold. In the second approach, a graph adjacency matrix is derived for the manifold, and the learned features are encouraged to be aligned with the principal components of this adjacency matrix. Experimental results on benchmark datasets are demonstrated, showing an obvious advantage of the proposed approaches when the training set is small.

Stochastic optimization makes it possible to track a slowly varying subspace underlying streaming data. By approximating local neighborhoods using affine subspaces, a slowly varying manifold can be efficiently tracked as well, even with corrupted and noisy data. The more the local neighborhoods, the better the approximation, but the higher the computational complexity. A multiscale approximation scheme is proposed, where the local approximating subspaces are organized in a tree structure. Splitting and merging of the tree nodes then allows efficient control of the number of neighbourhoods. Deviation (of each datum) from the learned model is estimated, yielding a series of statistics for anomaly detection. This framework extends the classical {\em changepoint detection} technique, which only works for one dimensional signals. Simulations and experiments highlight the robustness and efficacy of the proposed approach in detecting an abrupt change in an otherwise slowly varying low-dimensional manifold.

The Mathematics Anxiety of Bilingual Community College Students

Math anxiety levels and performance outcomes were compared for bilingual and monolingual community college Intermediate Algebra students attending a culturally diverse urban commuter college. Participants (N = 618, 250 men, 368 women; 361 monolingual, 257 bilingual) completed the Abbreviated Math Anxiety Scale (AMAS) and a demographics instrument. Bilingual and monolingual students reported comparable mean AMAS scores (20.6 and 20.7, respectively) and comparable proportions of math anxious individuals (50% and 48%, respectively). Factor analysis of AMAS scores, using principal component analysis by varimax rotation, yielded similar two-factor structures for both populations -- assessment and learning content -- accounting for 65.6% of the trace for bilingual AMAS scores. Statistically significant predictor variables for levels of math anxiety for the bilingual participants included (a) preparatory course enrollment (β = .236, p = .041) with those enrolled in prior preparatory courses scoring higher, (b) education major (β = .285, p = .018) with education majors scoring higher, and (c) business major (β = .252, p = .032) with business majors scoring higher. One statistically significant predictor variable emerged for monolingual students, gender (β = -.085, p = .001) with females ranking higher. Age, income, race, ethnicity, U.S. origin, science or health science majors did not emerge as statistically significant predictor variables for either group. Similarities between monolingual and bilingual participants included statistically significant negative linear correlations between AMAS scores and course grades for both bilingual (r = -.178, p = .017) and monolingual participants (r = -.203, p = .001). Differences included a statistically significant linear correlation between AMAS scores and final exam grades for monolingual participants only (r = -.253, p < .0009) despite no statistically significant difference in the strength the linear relationship of the AMAS scores and the final exam scores between groups, z = 1.35, p = .1756. The findings show that bilingual and monolingual students report math anxiety similarly and that math anxiety has similar associations with performance measures, despite differences between predictor variables. One of the first studies on the math anxiety of bilingual community college students, the results suggest recommendations for researchers and practitioners.

Migrant domestic workers, vulnerability and the law: immigration and employment laws in Cyprus and Spain

The intersection of gender, welfare and immigration regimes has been one of the main focus of a rich scholarship on paid domestic work in Europe. This article brings into the discussion the nexus of employment and immigration law regimes to reflect on the role of legal regulation in structuring and reducing the vulnerability of domestic workers. I analyse this nexus by looking at the cases of Cyprus and Spain, two states falling under the cluster of Southern Mediterranean welfare regimes, that share certain characteristics in terms of immigration regimes, but have substantially different employment law regulation models. The first part sketches the debate on the employment law regulation of domestic work. The second part starts by giving an overview of the immigration regimes of Cyprus and Spain in relation to migrant domestic workers and then proceeds to analyse the two countries’ models and substance of employment law regulation in domestic work. The comparison of these two divergent approaches informs the debate on how the legal regulation of domestic work should be best structured. In Spain there have been recent dynamic legislative changes in the employment law regulation of domestic work. The final part of the article traces these changes and reflects on why such processes have not taken place in Cyprus.

The Communicative Dimension and Security in Asia-Pacific: A communicative-viewing proposal for reform of the Japanese Intelligence Services

The postwar development of the Intelligence Services in Japan has been based on two contrasting models: the centralized model of the USA and the collegiality of UK, neither of which has been fully developed. This has led to clashes of institutional competencies and poor anticipation of threats towards national security. This problem of opposing models has been partially overcome through two dimensions: externally through the cooperation with the US Intelligence Service under the Treaty of Mutual Cooperation and Security; and internally though the pre-eminence in the national sphere of the Department of Public Safety. However, the emergence of a new global communicative dimension requires that a communicative-viewing remodeling of this dual model is necessary due to the increasing capacity of the individual actors to determine the dynamics of international events. This article examines these challenges for the Intelligence Services of Japan and proposes a reform based on this new global communicative dimension.

Commentary on: The Articles of the Pope's Bulle (1518)

One of the earliest examples of works printed by Richard Pynson, the King's Printer between 1508 and 1530, to make reference to the fact that the work in question was printed under the protection of the King. The royal printing privilege provided one of two different models for preventing the unauthorised reproduction of works after publication which prefigured the introduction of statutory copyright in the early eighteenth century.The commentary describes the early attitudes of the monarchy towards the regulation of the printing trade within England, and the exercise of the royal prerogative in granting printing privileges not just to the royal printer, but to other favoured subjects both in relation to individual works as well as to entire classes of work (with the latter more often referred to as printing patents).

Commentary on early Tudor printing privileges 1553: Totell's Printing Patent (1553)

One of the earliest examples of work printed by Richard Pynson, the King's Printer between 1508 and 1530, to make reference to the fact that the work in question was printed under the protection of the King. The royal printing privilege provided one of two different models for preventing the unauthorised reproduction of works after publication which prefigured the introduction of statutory copyright in the early eighteenth century.
The commentary describes the early attitudes of the monarchy towards the regulation of the printing trade within England, and the exercise of the royal prerogative in granting printing privileges not just to the royal printer, but to other favoured subjects both in relation to individual works as well as to entire classes of work (with the latter more often referred to as printing patents).

Extended Energy Vector Prediction of Ambisonically Reproduced Image Direction at Off-Centre Listening Positions

This paper presents an extension to the energy vector, well known in the Ambisonics literature, to improve its predictions of localisation at off-centre listening positions. In determining the source direction, a perceptual weight is assigned to each loudspeaker gain, taking into account the relative arrival times, levels, and directions of the loudspeaker signals. The proposed model is evaluated alongside the original energy vector and two binaural models through comparison with the results of recent perceptual studies. The extended version was found to provide results that were at least 50% more accurate than the second best predictor for two experiments involving off-centre listeners with first- and third-order Ambisonics systems.

Synchronized and controlled release of metformin hydrochloride/glipizide from elementary osmotic delivery

The combination of metformin hydrochloride (MTF) and glipizide (GLZ) is second-line medication for diabetes mellitus type 2 (DMT2). In the present study, elementary osmotic pump(EOP)tablet is designed to deliver the combination of MTF and GLZ in a sustained and synchronized manner. By analyzing different variables of the formulation, sodium hydrogen carbonate is introduced as pH modifier to improve the release of GLZ, while ethyl cellulose acts as release retardant to reduce the burst release phase of MTF. A two factor, three level face-centered central composite design (FCCD) is applied to investigate the impact of different factors on drug release profile. Compared with conventional tablets, the elementary osmotic pump (EOP) tablet demonstrates a controlled release behavior with relative bioavailability of 99.2% for MTF and 99.3% for GLZ. Data also shows EOP tablet is able to release MTF and GLZ in a synchronized and sustained manner both in vitro and in vivo

Temperature fields in a channel partially filled with a porous material under local thermal non-equilibrium condition-an exact solution

This work examines analytically the forced convection in a channel partially filled with a porous material and subjected to constant wall heat flux. The Darcy–Brinkman–Forchheimer model is used to represent the fluid transport through the porous material. The local thermal non-equilibrium, two-equation model is further employed as the solid and fluid heat transport equations. Two fundamental models (models A and B) represent the thermal boundary conditions at the interface between the porous medium and the clear region. The governing equations of the problem are manipulated, and for each interface model, exact solutions, for the solid and fluid temperature fields, are developed. These solutions incorporate the porous material thickness, Biot number, fluid to solid thermal conductivity ratio and Darcy number as parameters. The results can be readily used to validate numerical simulations. They are, further, applicable to the analysis of enhanced heat transfer, using porous materials, in heat exchangers.

Estudo Preliminar das Propriedades Psicométricas do Inventário de Lateralidade de Edinburgh numa Amostra de Adultos da População Portuguesa

Introdução: A lateralidade é a diferença na capacidade de controlo entre os dois lados do corpo. Os métodos utilizados para avaliar a lateralidade manual incluem a observação efetiva do uso do membro dominante ou a aplicação de inventários respondidos pelo próprio indivíduo avaliado. O Inventário de Lateralidade de Edinburgh (EHI) é o instrumento mais utilizado para avaliar a lateralidade manual. Apesar do seu uso amplo, em Portugal não existem estudos que avaliem a sua validade e fidedignidade. Objetivos: Estudar as propriedades psicométricas do Inventário de Lateralidade de Edinburgh numa amostra da população portuguesa. Métodos: A amostra é constituída por 290 pessoas (135 homens e 155 mulheres), com idades compreendidas entre os 18 e os 65 anos. Todos os participantes preencheram uma declaração de consentimento informado e uma bateria de testes neuropsicológicos Resultados: A média no EHI foi de 62,36 (DP = 38,00). Os resultados demonstraram que das seis variáveis sociodemográficas (idade, sexo, escolaridade, zona de residência, regiões e profissão) três apresentaram ter influência significativa nas pontuações do EHI: idade, zona de residência e regiões. A confiabilidade e a estabilidade temporal do EHI apresentaram resultados adequados. A análise fatorial confirmatória mostrou que o modelo não é melhor explicado por um fator. Para dois fatores o modelo continua a não ser adequado. Conclusão: Apesar de termos obtido uma boa consistência interna não nos é possível considerar este teste como o mais adequado para medir o constructo da lateralidade. / Introduction: The handedness is the difference in the control capacity between the two sides of the body. The methods used to evaluate the manual handedness include the effective observation of the use of dominant member or application of inventories answered by the person assessed. The Edinburgh Handedness Inventory (EHI) is the most used to evaluate manual handedness. Even though being widely used, in Portugal there are no studies that measure its validity and reliability. Objective: To study the psychometric properties of Edinburgh Handedness Inventory in a Portuguese sample. Methods: The sample consists of 290 people (135 men and 155 women), aged between 18 and 65 years. All participants filled an informed consent form and a battery of neuropsychological tests. Results: The average in EHI was 62.36 (SD = 38.00). The results showed that 3 of 6 sociodemographic variables showed significant influence in EHI scores. The reliability and temporal stability of EHI were adequate. Confirmatory factor analysis showed that the model is not better explained by one factor. A two-factor model was not also suitable. Conclusion: Even though we got a good internal consistency we cannot consider this test as the most appropriate for measuring the handedness construct.

Estudo Preliminar das Propriedades Psicométricas e dos Dados Normativos da Forma Geral das Matrizes Progressivas de Raven numa Amostra da Comunidade

Objetivo: Estudar as propriedades psicométricas e os dados normativos da Forma Geral das Matrizes Progressivas de Raven numa amostra da comunidade da população portuguesa. Método: A amostra é constituída por 697 pessoas (314 homens e 383 mulheres), com idades compreendidas entre os 12 e os 90 anos. Todos os participantes preencheram uma declaração de consentimento informado e uma bateria de testes neuropsicológicos, incluindo a Forma Geral das Matrizes Progressivas de Raven (FG-MPR), Teste de Memória de 15-Item de Rey, Escala de Autoavaliação de Ansiedade de Zung, Bateria de Avaliação Frontal e Figura Complexa de Rey. Resultados: A média na FG-MPR foi de 44,47 (DP = 10,78). Os resultados demonstraram que todas as variáveis sociodemográficas (idade, sexo, escolaridade, profissão, regiões e tipologia de áreas urbanas), exceto o estado civil, apresentaram ter influência significativa nas pontuações da FG-MPR. A confiabilidade e a estabilidade temporal da FG-MPR revelaram-se adequadas. A análise fatorial exploratória e confirmatória mostrou que o modelo para um fator não é adequado. Um modelo a quatro fatores continua a não ser adequado. Conclusão: Os dados do presente estudo sugerem que se trata de um instrumento com potencialidades na sua utilização junto da população portuguesa. / Purpose: To study the psychometric properties and date normative of the Raven’s Standard Progressive Matrices in a Portuguese community sample. Method: The sample consists of 697 people (314 men and 383 women), aged between 12 and 90 years. All participants filled an informed consent form and a battery of neuropsychological tests, which included Raven’s Standard Progressive Matrices (RSPM), Rey 15-Item Memory Test, Zung Self-Rating Anxiety Scale, Frontal Assessment Battery, and Rey Complex Figure Test. Results: The average in RSPM was 44.47 (SD = 10.78). The results showed that all of the sociodemographic variables (age, sex, education, profession, region, and typology of urban areas), with the exception of civil status, showed significant influence on RSPM scores. The reliability and temporal stability of RSPM were adequate. Exploratory and Confirmatory factor analysis showed that the model is not better explained by one factor. A two-factor model was not also suitable. Conclusion: The data from this study suggest that it is an instrument with potential for its use among the Portuguese population.

Combinative Aspects of Leadership Style and Emotional Intelligence

Purpose – The purpose of this paper is to examine whether the leader’s emotional intelligence influences the leader’s preferences for different ways of combining leadership behaviors (i.e. combinative aspects of leadership style). Design/methodology/approach – The authors used a hybrid design to collect the data to avoid common-method biases. The authors described a high-stress workplace in a vignette and asked participants to rank four styles of combining a task-oriented leadership (i.e. Pressure) statement and a socio-emotional leadership (i.e. Support) statement. The authors then asked participants to complete a Likert-scale based questionnaire on emotional intelligence. Findings – The authors found that leaders who prefer to provide Support immediately before Pressure have higher levels of emotional intelligence than do leaders who prefer the three other combinative styles. Leaders who prefer to provide Pressure and Support separately (i.e. provide Pressure 30 minutes after Support) have the lowest levels of emotional intelligence. Research limitations/implications – A key implicit assumption in the work is that leaders do not want to evoke negative emotions in followers. The authors did not take into account factors that influence leadership style which participating managers would be likely to encounter on a daily basis such as the relationship with the follower, the follower’s level of performance and work experience, the gender of the leader and the gender of the follower, the hierarchical levels of the leader and follower, and the followers’ preferred combinative style. The nature of the sample and the use of a hypothetical scenario are other limitations of the study. Practical implications – Providing leadership behaviors that are regarded as effective is necessary but not enough because the emotional impact of leadership behaviors appears to also depend on how the behaviors are configured. Originality/value – This is the first study to show that the emotional intelligence of leaders is related to their preferences for the manner in which they combine task and social leadership statements. Furthermore, two-factor theories of leadership propose that the effects of task and social leadership are additive. However, the findings show that the effects are interactive.

Cost-Efficient Spatial Placement of Water Retention Sites (WRS) to Reduce Sediment in Le Sueur River Watershed,MN

Thesis (Master's)--University of Washington, 2016-06

The role of metalinguistic awareness and of L2 proficiency in positive lexical transfer from English (L2) to German (L3) by French-speaking Quebeckers

Il a été avancé que des apprenants expérimentés développeraient des niveaux élevés de conscience métalinguistique (MLA), ce qui leur faciliterait l’apprentissage de langues subséquentes (p.ex., Singleton & Aronin, 2007). De plus, des chercheurs dans le domaine de l’acquisition des langues tierces insistent sur les influences positives qu’exercent les langues précédemment apprises sur l’apprentissage formel d’une langue étrangère (p.ex., Cenoz & Gorter, 2015), et proposent de délaisser le regard traditionnel qui mettait l’accent sur l’interférence à l’origine des erreurs des apprenants pour opter pour une vision plus large et positive de l’interaction entre les langues. Il a été démontré que la similarité typologique ainsi que la compétence dans la langue source influence tous les types de transfert (p.ex., Ringbom, 1987, 2007). Cependant, le défi méthodologique de déterminer, à la fois l’usage pertinent d’une langue cible en tant que résultat d’une influence translinguistique (p.ex., Falk & Bardel, 2010) et d’établir le rôle crucial de la MLA dans l’activation consciente de mots ou de constructions reliés à travers différentes langues, demeure. La présente étude avait pour but de relever ce double défi en faisant appel à des protocoles oraux (TAPs) pour examiner le transfert positif de l’anglais (L2) vers l’allemand (L3) chez des Québécois francophones après cinq semaines d’enseignement formel de la L3. Les participants ont été soumis à une tâche de traduction développée aux fins de la présente étude. Les 42 items ont été sélectionnés sur la base de jugements de similarité et d’imagibilité ainsi que de fréquence des mots provenant d’une étude de cognats allemands-anglais (Friel & Kennison, 2001). Les participants devaient réfléchir à voix haute pendant qu’ils traduisaient des mots inconnus de l’allemand (L3) vers le français (L1). Le transfert positif a été opérationnalisé par des traductions correctes qui étaient basées sur un cognat anglais. La MLA a été mesurée par le biais du THAM (Test d’habiletés métalinguistiques) (Pinto & El Euch, 2015) ainsi que par l’analyse des TAPs. Les niveaux de compétence en anglais ont été établis sur la base du Michigan Test (Corrigan et al., 1979), tandis que les niveaux d’exposition ainsi que l’intérêt envers la langue et la culture allemandes ont été mesurés à l’aide d’un questionnaire. Une analyse fine des TAPs a révélé de la variabilité inter- et intra-individuelle dans l’activation consciente du vocabulaire en L2, tout en permettant l’identification de niveaux distincts de prise de conscience. Deux modèles indépendants de régressions logistiques ont permis d’identifier les deux dimensions de MLA comme prédicteurs de transfert positif. Le premier modèle, dans lequel le THAM était la mesure exclusive de MLA, a déterminé cette dimension réflexive comme principal prédicteur, suivie de la compétence en anglais, tandis qu’aucune des autres variables indépendantes pouvait prédire le transfert positif de l’anglais. Dans le second modèle, incluant le THAM ainsi que les TAPs comme mesures complémentaires de MLA, la dimension appliquée de MLA, telle que mesurée par les TAPs, était de loin le prédicteur principal, suivie de la dimension réflexive, telle que mesurée par le THAM, tandis que la compétence en anglais ne figurait plus parmi les facteurs ayant une influence significative sur la variable réponse. Bien que la verbalisation puisse avoir influencé la performance dans une certaine mesure, nos observations mettent en évidence la contribution précieuse de données introspectives comme complément aux résultats basés sur des caractéristiques purement linguistiques du transfert. Nos analyses soulignent la complexité des processus métalinguistiques et des stratégies individuelles, ce qui reflète une perspective dynamique du multilinguisme (p.ex., Jessner, 2008).

Conception, fabrication et caractérisation d'un biocapteur SPR à base de guides d'ondes photoniques sur substrat de verre

Résumé : Malgré le nombre croissant de capteurs dans les domaines de la chimie et la biologie, il reste encore à étudier en profondeur la complexité des interactions entre les différentes molécules présentes lors d’une détection à l’interface solide-liquide. Dans ce cadre, il est de tout intérêt de croiser différentes méthodes de détection afin d’obtenir des informations complémentaires. Le principal objectif de cette étude est de dimensionner, fabriquer et caractériser un détecteur optique intégré sur verre basé sur la résonance plasmonique de surface, destiné à terme à être combiné avec d’autres techniques de détection, dont un microcalorimètre. La résonance plasmonique de surface est une technique reconnue pour sa sensibilité adaptée à la détection de surface, qui a l’avantage d’être sans marquage et permet de fournir un suivi en temps réel de la cinétique d’une réaction. L’avantage principal de ce capteur est qu’il a été dimensionné pour une large gamme d’indice de réfraction de l’analyte, allant de 1,33 à 1,48. Ces valeurs correspondent à la plupart des entités biologiques associées à leurs couches d’accroche dont les matrices de polymères, présentés dans ce travail. Étant donné que beaucoup d’études biologiques nécessitent la comparaison de la mesure à une référence ou à une autre mesure, le second objectif du projet est d’étudier le potentiel du système SPR intégré sur verre pour la détection multi-analyte. Les trois premiers chapitres se concentrent sur l’objectif principal du projet. Le dimensionnement du dispositif est ainsi présenté, basé sur deux modélisations différentes, associées à plusieurs outils de calcul analytique et numérique. La première modélisation, basée sur l’approximation des interactions faibles, permet d’obtenir la plupart des informations nécessaires au dimensionnement du dispositif. La seconde modélisation, sans approximation, permet de valider le premier modèle approché et de compléter et affiner le dimensionnement. Le procédé de fabrication de la puce optique sur verre est ensuite décrit, ainsi que les instruments et protocoles de caractérisation. Un dispositif est obtenu présentant des sensibilités volumiques entre 1000 nm/RIU et 6000 nm/RIU suivant l’indice de réfraction de l’analyte. L’intégration 3D du guide grâce à son enterrage sélectif dans le verre confère au dispositif une grande compacité, le rendant adapté à la cointégration avec un microcalorimètre en particulier. Le dernier chapitre de la thèse présente l’étude de plusieurs techniques de multiplexage spectral adaptées à un système SPR intégré, exploitant en particulier la technologie sur verre. L’objectif est de fournir au moins deux détections simultanées. Dans ce cadre, plusieurs solutions sont proposées et les dispositifs associés sont dimensionnés, fabriqués et testés.