Matching Items (3)
Filtering by
- Creators: Li, Baoxin
- Member of: ASU Electronic Theses and Dissertations
Description
Deep learning architectures have been widely explored in computer vision and have
depicted commendable performance in a variety of applications. A fundamental challenge
in training deep networks is the requirement of large amounts of labeled training
data. While gathering large quantities of unlabeled data is cheap and easy, annotating
the data is an expensive process in terms of time, labor and human expertise.
Thus, developing algorithms that minimize the human effort in training deep models
is of immense practical importance. Active learning algorithms automatically identify
salient and exemplar samples from large amounts of unlabeled data and can augment
maximal information to supervised learning models, thereby reducing the human annotation
effort in training machine learning models. The goal of this dissertation is to
fuse ideas from deep learning and active learning and design novel deep active learning
algorithms. The proposed learning methodologies explore diverse label spaces to
solve different computer vision applications. Three major contributions have emerged
from this work; (i) a deep active framework for multi-class image classication, (ii)
a deep active model with and without label correlation for multi-label image classi-
cation and (iii) a deep active paradigm for regression. Extensive empirical studies
on a variety of multi-class, multi-label and regression vision datasets corroborate the
potential of the proposed methods for real-world applications. Additional contributions
include: (i) a multimodal emotion database consisting of recordings of facial
expressions, body gestures, vocal expressions and physiological signals of actors enacting
various emotions, (ii) four multimodal deep belief network models and (iii)
an in-depth analysis of the effect of transfer of multimodal emotion features between
source and target networks on classification accuracy and training time. These related
contributions help comprehend the challenges involved in training deep learning
models and motivate the main goal of this dissertation.
depicted commendable performance in a variety of applications. A fundamental challenge
in training deep networks is the requirement of large amounts of labeled training
data. While gathering large quantities of unlabeled data is cheap and easy, annotating
the data is an expensive process in terms of time, labor and human expertise.
Thus, developing algorithms that minimize the human effort in training deep models
is of immense practical importance. Active learning algorithms automatically identify
salient and exemplar samples from large amounts of unlabeled data and can augment
maximal information to supervised learning models, thereby reducing the human annotation
effort in training machine learning models. The goal of this dissertation is to
fuse ideas from deep learning and active learning and design novel deep active learning
algorithms. The proposed learning methodologies explore diverse label spaces to
solve different computer vision applications. Three major contributions have emerged
from this work; (i) a deep active framework for multi-class image classication, (ii)
a deep active model with and without label correlation for multi-label image classi-
cation and (iii) a deep active paradigm for regression. Extensive empirical studies
on a variety of multi-class, multi-label and regression vision datasets corroborate the
potential of the proposed methods for real-world applications. Additional contributions
include: (i) a multimodal emotion database consisting of recordings of facial
expressions, body gestures, vocal expressions and physiological signals of actors enacting
various emotions, (ii) four multimodal deep belief network models and (iii)
an in-depth analysis of the effect of transfer of multimodal emotion features between
source and target networks on classification accuracy and training time. These related
contributions help comprehend the challenges involved in training deep learning
models and motivate the main goal of this dissertation.
ContributorsRanganathan, Hiranmayi (Author) / Sethuraman, Panchanathan (Thesis advisor) / Papandreou-Suppappola, Antonia (Committee member) / Li, Baoxin (Committee member) / Chakraborty, Shayok (Committee member) / Arizona State University (Publisher)
Created2018
Description
Interpersonal strain is linked with depressive symptoms in middle-aged adults. Self-compassion is an emerging resilience construct that may be advantageous in navigating relationship strain by helping individuals respond to emotions in a kind and nonjudgmental way. Although theory and empirical evidence suggests that self-compassion is protective against the impact of stress on mental health outcomes, many studies have not investigated how self-compassion operates in the context of relationship strain. In addition, few studies have examined psychological or physiological mechanisms by which self-compassion protects against mental health outcomes, depression in particular. Thus, this study examined 1) the extent to which trait self-compassion buffers the relation between family strain and depressive symptoms, and 2) whether these buffering effects are mediated by hope and inflammatory processes (IL-6) in a sample of 762 middle-aged, community-dwelling adults. Results from structural equation models indicated that family strain was unrelated to depressive symptoms and the relation was not moderated by self-compassion. Hope, but not IL-6, mediated the relation between family strain and depressive symptoms and the indirect effect was not conditional on levels of self-compassion. Taken together, the findings suggest that family strain may lead individuals to experience less hope and subsequent increases in depressive symptoms, and further, that a self-compassionate attitude does not affect this relation. Implications for future self-compassion interventions are discussed.
ContributorsMistretta, Erin (Author) / Davis, Mary C. (Thesis advisor) / Karoly, Paul (Committee member) / Infurna, Frank (Committee member) / Arizona State University (Publisher)
Created2019
Description
Lifespan psychological perspectives have long suggested the context in which individuals live having the potential to shape the course of development across the adult lifespan. Thus, it is imperative to examine the role of both the objective and subjective neighborhood context in mitigating the consequences of lifetime adversity on mental and physical health. To address the research questions, data was used from a sample of 362 individuals in midlife who were assessed on lifetime adversity, multiple outcomes of mental and physical health and aspects of the objective and subjective neighborhood. Results showed that reporting more lifetime adversity was associated with poorer mental and physical health. Aspects of the objective and subjective neighborhood, such as green spaces moderated these relationships. The discussion focuses on potential mechanisms underlying why objective and subjective indicators of the neighborhood are protective against lifetime adversity.
ContributorsStaben, Omar E (Author) / Infurna, Frank J. (Thesis advisor) / Luthar, Suniya S. (Committee member) / Grimm, Kevin J. (Committee member) / Arizona State University (Publisher)
Created2019