No description available.

<p>Domain adaptation is one of the hottest directions in solving annotation insufficiency problem of deep learning. General domain adaptation is not consistent with the practical scenarios in the industry. In this thesis, we focus on two concerns as below.</p> <p>  </p> <p>  First is that labeled data are generally collected from multiple domains. In other words, multi-source adaptation is a more common situation. Simply extending these single-source approaches to the multi-source cases could cause sub-optimal inference, so specialized multi-source adaptation methods are essential. The main challenge in the multi-source scenario is a more complex divergence situation. Not only the divergence between target and each source plays a role, but the divergences among distinct sources matter as well. However, the significance of maintaining consistency among multiple sources didn't gain enough attention in previous work. In this thesis, we propose an Enhanced Consistency Multi-Source Adaptation (EC-MSA) framework to address it from three perspectives. First, we mitigate feature-level discrepancy by cross-domain conditional alignment, narrowing the divergence between each source and target domain class-wisely. Second, we enhance multi-source consistency via dual mix-up, diminishing the disagreements among different sources. Third, we deploy a target distilling mechanism to handle the uncertainty of target prediction, aiming to provide high-quality pseudo-labeled target samples to benefit the previous two aspects. Extensive experiments are conducted on several common benchmark datasets and demonstrate that our model outperforms the state-of-the-art methods.</p> <p>  </p> <p>  Second is that data privacy and security is necessary in practice. That is, we hope to keep the raw data stored locally while can still obtain a satisfied model. In such a case, the risk of data leakage greatly decreases. Therefore, it is natural for us to combine the federated learning paradigm with domain adaptation. Under the source-private setting, the main challenge for us is to expose information from the source domain to the target domain while make sure that the communication process is safe enough. In this thesis, we propose a method named Fourier Transform-Assisted Federated Domain Adaptation (FTA-FDA) to alleviate the difficulties in two ways. We apply Fast Fourier Transform to the raw data and transfer only the amplitude spectra during the communication. Then frequency space interpolations between these two domains are conducted, minimizing the discrepancies while ensuring the contact of them and keeping raw data safe. What's more, we make prototype alignments by using the model weights together with target features, trying to reduce the discrepancy in the class level. Experiments on Office-31 demonstrate the effectiveness and competitiveness of our approach, and further analyses prove that our algorithm can help protect privacy and security.</p>

No description available.

No description available.

No description available.

Athletics has been regarded as a means of encouraging youth to build character, discipline, and develop healthy habits. However, literature has emerged that asserts athletics do not prevent deviant behaviors, but instead, influence one to commit deviant acts. As such, this research examined effects of athletics on the commission of deviant behaviors via learning and techniques of neutralization theories. Subjects for this project included 325 college students from a southern regional university. Data were generated through the use of a self-report questionnaire, which measured variables pertaining to self-reported deviant behaviors including perceptions of peer deviance, neutralizing indicators, and sports participation. The findings suggest some support for each theoretical model, differential association and techniques of neutralization. Both theoretical models were supported, in general, with learning theory having the most support. When participation in sporting activity was considered, however, the results consistently showed no effect on various types of self-reported deviant behavior.

This study explored the psychological and sociological motivations of adult female and male obstacle course racers. A qualitative case study approach was used to explore the views, experiences, and motivations of obstacle course racing (OCR) participants. Descriptive statistics and cross tabulation was used to interpret responses to the 297 online questionnaires. A content analysis approach was used to analyze the qualitative data gathered from three focus groups with a total of 20 obstacle course racers. Three theories formed the basis of the study: Self-Determination Theory (SDT), Achievement Goal Theory (AGT), and Social Leaning Theory (SLT). Overall, findings supported previous research regarding motivations to participate in adventure racing and extreme sports. Individuals were guided more by intrinsic motives than extrinsic motives. Important motivations for obstacle course racers included the camaraderie among participants, connecting and socializing with other like-minded people, having fun, and having a physical challenge that allowed them to progress and keep on track with their health goals. Obstacle course racing was perceived as positively impacting participants’ health, mental wellness and their confidence in their physical abilities as well as in other areas of their lives. Findings from this study may inform future interventions to increase participation in OCR or to increase overall physical activity among adults by building on camaraderie, social connection, enjoyment, and self-efficacy.

The current study sought to explore the effects of eating behaviors on mental health and the internalization of attitudes toward one's appearance. This was performed through a survey distributed through a global online market research firm, Dynata, and gathered 495 eligible participants. Of those, 78 represented the plant-based group, and 417 represented the non-plant-based group. Each completed the survey containing sections with the Mizes Anorectic Cognitions Scale (MACS) to assess if they were at low or high risk for having/developing an eating disorder, a section determining if participants were plant-based (defined as a regimen that encourages whole, plant-based foods and discourages meats, dairy products, and eggs as well as all refined and processed foods (Tuso et al., 2013)) or not, and asking what their perceived benefits were from their plant-based (or lack of plant-based) diet, a section on the Sociocultural Attitudes Towards Appearance (SATAQ-4), and finally, the Depression, Anxiety, and Stress Scale (DASS). The status of participants being plant-based or non-plant-based was analyzed as this study sought to explore past research that found plant-based diets to be physically and mentally beneficial (Beezhold et al., 2014; Benefits of Plant-Based Diets, 2021; Daneshzad et al., 2019). A series of statistical tests were conducted on SPSS 28 to analyze which groups (high risk for E.D. and plant-based, high risk for E.D. and non-plant-based, low risk for E.D. and plant-based, or low risk for E.D. and non-plant-based) were statistically significant compared to one another. The findings revealed the plant-based group to contain higher percentage of high risk for eating disorder participants. The plant-based group, regardless of high risk, was negatively associated with higher levels of scores on four out of the five sections including internalization of attitudes towards appearance, weight-based pressures from family, weight-based pressures from peers, and higher levels of reported depression, anxiety, and stress. Pressures felt from the media did not display a statistically significant level of difference between any of the high/low risk and plant-based to high/low risk and non-plant-based. The results were interpreted using social learning theory, which proposes that humans have evolved an advanced capacity for observational learning, enabling them to acquire knowledge, attitudes, values, emotional proclivities, and competencies through information conveyed by a rich variety of actual and symbolic models (Bandura, 2002). This allowed for cause and effects to be hypothesized for why the plant-based group was negatively associated with worse mental health and internalization of attitudes towards appearance. Among these hypothesized causes included participants adopting a plant-based diet due to its growing social media popularity, users learning from observation and leading to aquired knowledge, attitudes, values, and beliefs on the diet. Furthermore, those with an obsession of clean-eating could have led many high-risk for E.D. participants to fulfill their internalized thin-ideal and pressure from family and peers regarding appearance through this popular diet they have observed through social media, which would be consistent with previous studies (Holmgren, 2017; Stewart & Ogden, 2020). Limitations include the small sample size of plant-based dieters without equal representation of low to high risk for E.D.s, along with the limitation on not knowing the reasoning why each participant is plant-based (ethical, environmental, health, diet purposes, or due to social learning and popularity) nor for how long they have adhered to this lifestyle. Future research should expand this study to more locations, analyze for differences based on age groups, and build upon the current study to allow for more generalizability.

Bien qu'ils soient capables de représenter des fonctions très complexes, les réseaux de neurones profonds sont entraînés à l'aide de variations autour de la descente de gradient, un algorithme qui est basé sur une simple linéarisation de la fonction de coût à chaque itération lors de l'entrainement. Dans cette thèse, nous soutenons qu'une approche prometteuse pour élaborer une théorie générale qui expliquerait la généralisation des réseaux de neurones, est de s'inspirer d'une analogie avec les modèles linéaires, en étudiant le développement de Taylor au premier ordre qui relie des pas dans l'espace des paramètres à des modifications dans l'espace des fonctions. Cette thèse par article comprend 3 articles ainsi qu'une bibliothèque logicielle. La bibliothèque NNGeometry (chapitre 3) sert de fil rouge à l'ensemble des projets, et introduit une Interface de Programmation Applicative (API) simple pour étudier la dynamique d'entrainement linéarisée de réseaux de neurones, en exploitant des méthodes récentes ainsi que de nouvelles accélérations algorithmiques. Dans l'article EKFAC (chapitre 4), nous proposons une approchée de la Matrice d'Information de Fisher (FIM), utilisée dans l'algorithme d'optimisation du gradient naturel. Dans l'article Lazy vs Hasty (chapitre 5), nous comparons la fonction obtenue par dynamique d'entrainement linéarisée (par exemple dans le régime limite du noyau tangent (NTK) à largeur infinie), au régime d'entrainement réel, en utilisant des groupes d'exemples classés selon différentes notions de difficulté. Dans l'article NTK alignment (chapitre 6), nous révélons un effet de régularisation implicite qui découle de l'alignement du NTK au noyau cible, au fur et à mesure que l'entrainement progresse. / Despite being able to represent very complex functions, deep artificial neural networks are trained using variants of the basic gradient descent algorithm, which relies on linearization of the loss at each iteration during training. In this thesis, we argue that a promising way to tackle the challenge of elaborating a comprehensive theory explaining generalization in deep networks, is to take advantage of an analogy with linear models, by studying the first order Taylor expansion that maps parameter space updates to function space progress. This thesis by publication is made of 3 papers and a software library. The library NNGeometry (chapter 3) serves as a common thread for all projects, and introduces a simple Application Programming Interface (API) to study the linearized training dynamics of deep networks using recent methods and contributed algorithmic accelerations. In the EKFAC paper (chapter 4), we propose an approximate to the Fisher Information Matrix (FIM), used in the natural gradient optimization algorithm. In the Lazy vs Hasty paper (chapter 5), we compare the function obtained while training using a linearized dynamics (e.g. in the infinite width Neural Tangent Kernel (NTK) limit regime), to the actual training regime, by means of examples grouped using different notions of difficulty. In the NTK alignment paper (chapter 6), we reveal an implicit regularization effect arising from the alignment of the NTK to the target kernel as training progresses.

No description available.