Matching Items (12,990)
Filtering by
- Creators: Arizona State University
Description
Pushing the artificial intelligence frontier to resource-constrained edge nodes for edge intelligence is nontrivial. This dissertation provides a comprehensive study of optimization-based meta-learning algorithms to build a theoretic foundation of edge intelligence, with the focus on two topics: 1) model-based reinforcement learning (RL); 2) distributed edge learning. Under this common theme, this study is broadly organized into two parts. The first part studies meta-learning algorithms for model-based RL. First, the fundamental limit of model learning is explored for linear time-varying systems, using a two-step meta-learning algorithm with an episodic block model. A comprehensive non-asymptotic analysis of the sample complexity is provided, where a two-scale martingale small-ball approach is devised to address the challenges in sample correlation and small sample sizes. Next, policy learning of offline RL in general Markov decision processes is explored. To tackle the challenges therein, e.g., value overestimation and possibly poor quality of offline datasets, a model-based offline Meta-RL approach with regularized policy optimization is proposed, by learning a meta-model for task inference and a meta-policy for safe exploration of out-of-distribution state-actions. The second part investigates meta-learning algorithms for distributed edge learning. First, the general edge supervised learning is considered, where the edge node aims to quickly learn a good model with limited samples. A platform-aided collaborative learning framework is proposed to learn a model initialization via federated meta-learning across multiple nodes, which is transferred to target nodes for fine-tuning. Then, a channel gating module is introduced to select important channels of backbone networks for efficient local computation. A novel federated meta-learning approach is developed to learn meta-initializations for backbone networks and gating modules, from which a task-specific channel gated network is quickly adapted. Taking one step further, the continual edge learning is investigated in the context of online meta-learning, where each node has a sequence of online tasks. A multi-agent online meta-learning framework is developed to accelerate the task-average performance in a single node under limited communication among neighbors, through the lens of distributed online convex optimization. Building on distributed online gradient descent with gradient tracking, the optimal task-average regret is achieved at a faster rate.
ContributorsLin, Sen (Author) / Zhang, Junshan JZ (Thesis advisor) / Ying, Lei LY (Thesis advisor) / Bertsekas, Dimitri DB (Committee member) / Nedich, Angelia AN (Committee member) / Wang, Weina WW (Committee member) / Arizona State University (Publisher)
Created2021
Description
A general review of film growth with various mechanisms is given. Additives and their potential effects on film properties are also discussed. Experimental light-induced aluminum (Al) plating tool design is discussed. Light-induced electroplating of Al as the front electrode on the n-type emitter of silicon (Si) solar cells is proposed as a substitute for screen-printed Silver (Ag). The advantages and disadvantages of Al over copper (Cu) as a suitable Ag replacement are examined. Optimization of the power given to a green laser for silicon nitride (SiNx) anitreflection coating patterning is performed. Laser damage and contamination removal conditions on post-patterned cell surfaces are identified. Plating and post-annealing temperature effects on Al morphology and film resistivity are explored. Morphology and resistivity improvement of the Al film are also investigated through several plating additives. The lowest resistivity of 3.1 µΩ-cm is given by nicotinic acid. Laser induced damage to the cell emitter experimentally limits the contact resistivity between light-induced Al and Si to approximately 69 mΩ-cm2. Phosphorus pentachloride (PCl5) is introduced into the plating bath and improved the the contact resistivity between light induced Al and Si to a range of 0.1-1 mΩ-cm2. Secondary ion mass spectroscopy (SIMS) was performed on a film deposited with PCl5 and showed a phosphorus peak, indicating emitter phosphorus concentration may be the reason for the low contact resistivity between light-induced Al and Si. SEM also shows that PCl5 improves Al film density and plating throwing power. Post plating annealing performed at a temperature of 500°C allows Al to spike through the thin n-type emitter causing cell failure. Atmospheric moisture causes poor process reproducibility.
ContributorsRicci, Lewis (Author) / Tao, Meng (Thesis advisor) / Goryll, Michael (Committee member) / Kozicki, Michael (Committee member) / Yu, Hongbin (Committee member) / Arizona State University (Publisher)
Created2021
Description
Structural/system health monitoring (SHM) and prognostic health management (PHM) are vital techniques to ensure engineering system reliability and safety during the service. As multi-functionality and enhanced performance are in demand, modern engineering systems including aerospace, mechanical, and civil applications have become more complex. The constituent and architectural complexity, and multisource sensing sources in modern engineering systems may limit the monitoring capabilities of conventional approaches and require more advanced SHM/PHM techniques. Therefore, a hybrid methodology that incorporates information fusion, nondestructive evaluation (NDE), machine learning (ML), and statistical analysis is needed for more effective damage diagnosis/prognosis and system safety management.This dissertation presents an automated aviation health management technique to enable proactive safety management for both aircraft and national airspace system (NAS). A real-time, data-driven aircraft safety monitoring technique using ML models and statistical models is developed to enable an early-stage upset detection capability, which can improve pilot’s situational awareness and provide a sufficient safety margin. The detection accuracy and computational efficiency of the developed monitoring techniques is validated using commercial unlabeled flight data recorder (FDR) and reported accident FDR dataset. A stochastic post-upset prediction framework is developed using a high-fidelity flight dynamics model to predict the post-impacts in both aircraft and air traffic system. Stall upset scenarios that are most likely occurred during loss of control in-flight (LOC-I) operation are investigated, and stochastic flight envelopes and risk region are predicted to quantify their severities.
In addition, a robust, automatic damage diagnosis technique using ultrasonic Lamb waves and ML models is developed to effectively detect and classify fatigue damage modes in composite structures. The dispersion and propagation characteristics of the Lamb waves in a composite plate are investigated. A deep autoencoder-based diagnosis technique is proposed to detect fatigue damage using anomaly detection approach and automatically extract damage sensitive features from the waves. The patterns in the features are then further analyzed using outlier detection approach to classify the fatigue damage modes. The developed diagnosis technique is validated through an in-situ fatigue tests with periodic active sensing.
The developed techniques in this research are expected to be integrated with the existing safety strategies to enhance decision making process for improving engineering system safety without affecting the system’s functions.
ContributorsLee, Hyunseong (Author) / Chattopadhyay, Aditi (Thesis advisor) / Liu, Yongming (Committee member) / Papandreou-Suppappola, Antonia (Committee member) / Fard, Masoud Yekani (Committee member) / Tang, Pingbo (Committee member) / Campbell, Angela (Committee member) / Arizona State University (Publisher)
Created2021
Description
Our children come to school every day to learn, participate, and prepare for what the future will bring. Others come to school to find refuge and help from those who dedicate their lives to ensure they are well and safe. They come with their minds filled with hopes and dreams, while others walk around the hallways with their hearts filled with despair and uncertainty. Despite collaborative district efforts and improvements in student services, students continue to experience trauma related symptoms and other mental disorders at disconcerting rates. The school district reports that approximately 98% of students have experienced traumatic episodes and half of these students presented with significant distress from symptoms of Post-Traumatic Stress Disorder (Loudenback, 2016). At this school, approximately 25% of the student body has been referred, identified and treated for socio-emotional difficulties. These rates are often higher in students with learning disabilities participating in different academic programs. This action research study was conducted to evaluate how and to what extent does implementation of a resilience-based curriculum affect students’ resilience, Posttraumatic Stress Disorder (PTSD) symptoms, attitudes toward school and efficacy for coping. This project was implemented over ten consecutive weeks in an urban middle school in East Los Angeles to a group of twenty students in special education. The intervention consists of ten modules each with activities and strategies designed to raise the students’ resilience and overall well-being. Resilience Theory and Social Cognitive Theory provide the framework for understanding the problem of practice and informing the intervention. Research along with professional observations regarding the vulnerability of students in special education coupled with the lack of evidence-based practices that assist in their emotional development inspired this project. This action research relied on an explanatory sequential design where qualitative results explained and supported the results from the quantitative data. Following the explanatory design, quantitative data was collected analyzed followed by qualitative data upon completion of the intervention. Data collected from web-based surveys and focus groups demonstrate that their participation in the resilience-based intervention increased their resilience, more specifically self-efficacy and problem solving skills while reducing PTSD symptoms. Results also showed students improved their attitudes toward school and ability to cope with stress. Quantitative and qualitative data merging, interpretation, and relation to both theory and research are discussed along with the study’s limitations, implication for research and practice, and concluding thoughts.
ContributorsDussan, Francisco Jose (Author) / Gee, Elizabeth (Thesis advisor) / Wolf, Leigh (Committee member) / Elsasser, Jim (Committee member) / Arizona State University (Publisher)
Created2021
Description
Approximately 71% of the great lakes, lakes, reservoirs, and ponds, together with 51% of rivers and streams assessed in the US are impaired or threatened by pollution or do not meet the minimum water quality requirements. Pathogens, sediments, and nutrients are leading causes of impairment, with agriculture being a top source of pollution. Agricultural pollution has become a global concern overtaking urban contamination as the major factor of inland and coastal waters degradation in many parts of the world. High-yielding crop production has been achieved by the intensive use of inorganic fertilizers that are mainly composed of Nitrogen (N) and Phosphorus (P). N and P are essential nutrients for ecosystem structure, processes, and functions. However, N and P in excess can be problematic to the environment. One of the major impacts of the increasing amount of these nutrients in the environment is the global expansion of harmful algal blooms (HABs). Major agricultural nutrient pollution sources and climate change can exacerbate these risks. This dissertation aims to guide future policies to mitigate issues linked to excess nutrient loads in the U.S. by evaluating the impact of climate change on nutrient loads and assessing the environmental impact as well as the spatial patterns of one of the major agricultural sources of nutrient pollution - Concentrated Animal Feeding Operations (CAFOs). Specifically, I first investigated the impact of bias correction techniques when modeling mid-century nutrient loads in a watershed heavily impacted by CAFOs. Second, I evaluated the role of CAFOs in land use change and subsequent environmental degradation of the surrounding environment. Finally, I assessed the spatial organization of CAFOs and its links to water quality conditions. The findings revealed unique insights for future nutrient management strategies in the U.S.
ContributorsMiralha, Lorrayne (Author) / Muenich, Rebecca L. (Thesis advisor) / Garcia, Margaret (Committee member) / Xu, Tianfang (Committee member) / Myint, Soe W. (Committee member) / Arizona State University (Publisher)
Created2021
Description
Modern communication systems call for state-of-the-art links that offer almost idealistic performance. This requirement had pushed the technological world to pursue communication in frequency bands that were almost incomprehensible back when the first series of cordless cellphones were invented. These requirements have impacted everything from civilian requirements, space, medical diagnostics to defense technologies and have ushered in a new era of advancements. This work presents a new and novel approach towards improving the conventional phased array systems. The Intelligent Phase Shifter (IPS) offers phase tracking and discrimination solutions that currently plague High-Frequency wireless systems. The proposed system is implemented on (CMOS) process node to better scalability and reduce the overall power dissipated. A tracking system can discern Radio Frequency (RF) Signals’ phase characteristics using a double-balanced mixer. A locally generated reference signal is then matched to the phase of the incoming receiver using a fully modular yet continuous complete 360ᵒ phase shifter that alters the phase of the local reference and matches the phase with that of an incoming RF reference. The tracking is generally two control voltages that carry In-phase and Quadrature-phase information. These control signals offer the capability of controlling similar devices when placed in an array and eliminating any ambiguity that might occur due to in-band interference.
ContributorsLakshminarasimhaiah Rajendra, Yashas (Author) / Zeinolabedinzadeh, Saeed (Thesis advisor) / Trichopoulos, Georgios (Committee member) / Aberle, James (Committee member) / Arizona State University (Publisher)
Created2021
Description
“(Dis)Locating the Sensual: Black Queer Placemaking in Brooklyn, New York” investigates the impact that gentrification has on Black queer subject formations and Black queer public culture. My research explores the interplay between social oppressions, the Black quotidian queer body, and lived sensations within two Black queer bars located in the epicenter of white middle class gentrification in Brooklyn, New York: Langston’s Brooklyn and Happiness Lounge. In doing so, my project expands Western conceptions of space while charging feminist and queer theories to explore interpersonal and personal dimensions of lived experiences that are conditioned by modes of normalization set by white supremacy. I use archival research and ethnographic methods to explore the way that Black queer people utilize space and the spatial dimensions that happen in and across the spaces they regularly occupy. I also collect and examine building information, such as the owners’ respective rental agreements, building permits, documented building violation(s), and incurred fees by the owners to understand who owns the land, who manages the properties, and the role of the state in regulating space. Additionally, “(Dis)Locating the Sensual” analyzes three analytic memos from my ethnographic fieldnotes including desire, spatial performance, and sensations to apprehend the implications of performance on a Black queer sense of place. Taken together, this data renders a complex picture of Black queer place-making that both resists and exceeds the structural constraints of racial capitalist expansion. My work both dialogues with and contributes to fields that are rarely drawn into conversation: Urban geography and Black queer studies. By analyzing sensations, nostalgia, and atmosphere within Langston’s and Happiness Lounge, I chart the ways in which gentrification continues to displace physical Black queer social spaces and impact the atmospheres and sensations that are unique to their vanishing social spaces. I introduce Black queer spatiality as a method that is informed by tracing Black LGBT spatial sensations and atmospheres; this analytic enables the linking of physical spatio-historical processes of extraction to the sensual geographic experiences that are emplaced in Black queer social spaces.
ContributorsMillhouse, Ricardo (Author) / Bailey, Marlon M (Thesis advisor) / Shabazz, Gregg R (Committee member) / Fonow, Mary M (Committee member) / McHugh, Kevin (Committee member) / Arizona State University (Publisher)
Created2021
Description
This project features three new pieces for oboe commissioned from three different composers. Each piece explores styles and/or instrumentations that are less common in the current body of repertoire. These pieces are Scenes for Charlie by Bryan Kennard, Love’s Last Gift by Thomas Juneau, and But Joy Comes in the Morning by William Brusick. A performance guide has been included for each piece, providing tips and suggestions for musicians wanting to perform these pieces in the future. In addition to the performance guide, each composer answered a list of interview questions to provide background information and give insight into their compositional process. Accompanying this document are recordings performed by the author.
ContributorsSummers, Season (Author) / Schuring, Martin (Thesis advisor) / Gardner, Joshua (Committee member) / Norton, Kay (Committee member) / Arizona State University (Publisher)
Created2021
Description
Monitoring a system for deviations from standard or reference behavior is essential for many data-driven tasks. Whether it is monitoring sensor data or the interactions between system elements, such as edges in a path or transactions in a network, the goal is to detect significant changes from a reference. As technological advancements allow for more data to be collected from systems, monitoring approaches should evolve to accommodate the greater collection of high-dimensional data and complex system settings. This dissertation introduces system-level models for monitoring tasks characterized by changes in a subset of system components, utilizing component-level information and relationships. A change may only affect a portion of the data or system (partial change). The first three parts of this dissertation present applications and methods for detecting partial changes. The first part introduces a methodology for partial change detection in a simple, univariate setting. Changes are detected with posterior probabilities and statistical mixture models which allow only a fraction of data to change. The second and third parts of this dissertation center around monitoring more complex multivariate systems modeled through networks. The goal is to detect partial changes in the underlying network attributes and topology. The contributions of the second and third parts are two non-parametric system-level monitoring techniques that consider relationships between network elements. The algorithm Supervised Network Monitoring (SNetM) leverages Graph Neural Networks and transforms the problem into supervised learning. The other algorithm Supervised Network Monitoring for Partial Temporal Inhomogeneity (SNetMP) generates a network embedding, and then transforms the problem to supervised learning. At the end, both SNetM and SNetMP construct measures and transform them to pseudo-probabilities to be monitored for changes. The last topic addresses predicting and monitoring system-level delays on paths in a transportation/delivery system. For each item, the risk of delay is quantified. Machine learning is used to build a system-level model for delay risk, given the information available (such as environmental conditions) on the edges of a path, which integrates edge models. The outputs can then be used in a system-wide monitoring framework, and items most at risk are identified for potential corrective actions.
ContributorsKasaei Roodsari, Maziar (Author) / Runger, George (Thesis advisor) / Escobedo, Adolfo (Committee member) / Pan, Rong (Committee member) / Shinde, Amit (Committee member) / Arizona State University (Publisher)
Created2021
Description
The purpose of this study was to explore the influence of the innovation, the Professional Learning Community-Orientation Modules (PLC-OM), on new teachers’ (NTs) attitudes towards and self-efficacy for PLCs and their self-efficacy and abilities as NTs. The school district in which this study took place did not have any support for NTs who entered their Professional Learning Communities (PLCs). The PLC-OM was designed to address the lack of support for PLCs, increasing NTs knowledge of PLCs and PLC skills and empowering them to act within collaborative networks. The literature review includes various, relevant studies from areas such as new teacher retention, specifically induction and mentoring programs, NT collaboration, and NT self-efficacy and past research around PLCs and online learning communities. The theory guiding this study includes sociocultural theory, social cognitive theory, and communities of practice. This mixed-methods action research study was conducted within southeastern Pennsylvania and included a total of 18 participants from elementary, middle, and high school. The innovation was implemented over a 13-week period with participants engaging in asynchronous and synchronous activities through Schoology, a learning management system. Participants completed pre- and post-innovation surveys and the Perceived New Teacher Growth Level Survey. Additionally, throughout the PLC-OM, NTs completed a Flipgrid introduction, discussion board responses, and PLC reflections. Flipgrid is a video recording platform that allows participants to create short videos and share with a group. They also engaged in virtual synchronous meetings as an entire cohort which were led by the researcher and focus-group interviews. Quantitative data was analyzed through descriptive statistics and a one sample t-test for the pre- and post-innovation surveys while the qualitative data was analyzed using a grounded theory approach, specifically the constant comparative method. All data was triangulated to confirm and corroborate findings.
Results suggested that the PLC-OM was beneficial for NTs and contributed to an increase in self-efficacy for PLCs and as NTs. NTs showed an increase in knowledge of PLCs and their PLC skills including interpersonal skills that can assist with collaboration. Additionally, the PLC-OM positively influenced NTs attitudes toward PLCs and their abilities as NTs. The discussion focuses on clarifying the following: the changes in NTs self-efficacy for PLCs and as NTs; the attitudes of NTs toward PLCs; and the influence of the PLC-OM design.
ContributorsForrest, Nicole (Author) / Markos, Amy (Thesis advisor) / Florio, Tammi (Committee member) / Farmakis, Heather (Committee member) / Arizona State University (Publisher)
Created2021