Matching Items (1,115)
Filtering by
- All Subjects: Computer Science
Description
Quantum computing holds the potential to revolutionize various industries by solving problems that classical computers cannot solve efficiently. However, building quantum computers is still in its infancy, and simulators are currently the best available option to explore the potential of quantum computing. Therefore, developing comprehensive benchmarking suites for quantum computing simulators is essential to evaluate their performance and guide the development of future quantum algorithms and hardware. This study presents a systematic evaluation of quantum computing simulators’ performance using a benchmarking suite. The benchmarking suite is designed to meet the industry-standard performance benchmarks established by the Defense Advanced Research Projects Agency (DARPA) and includes standardized test data and comparison metrics that encompass a wide range of applications, deep neural network models, and optimization techniques. The thesis is divided into two parts to cover basic quantum algorithms and variational quantum algorithms for practical machine-learning tasks. In the first part, the run time and memory performance of quantum computing simulators are analyzed using basic quantum algorithms. The performance is evaluated using standardized test data and comparison metrics that cover fundamental quantum algorithms, including Quantum Fourier Transform (QFT), Inverse Quantum Fourier Transform (IQFT), Quantum Adder, and Variational Quantum Eigensolver (VQE). The analysis provides valuable insights into the simulators’ strengths and weaknesses and highlights the need for further development to enhance their performance. In the second part, benchmarks are developed using variational quantum algorithms for practical machine learning tasks such as image classification, natural language processing, and recommendation. The benchmarks address several unique challenges posed by benchmarking quantum machine learning (QML), including the effect of optimizations on time-to-solution, the stochastic nature of training, the inclusion of hybrid quantum-classical layers, and the diversity of software and hardware systems. The findings offer valuable insights into the simulators’ ability to solve practical machine-learning tasks and pinpoint areas for future research and enhancement. In conclusion, this study provides a rigorous evaluation of quantum computing simulators’ performance using a benchmarking suite that meets industry-standard performance benchmarks.
ContributorsSathyakumar, Rajesh (Author) / Spanias, Andreas (Thesis advisor) / Sen, Arunabha (Thesis advisor) / Dasarathy, Gautam (Committee member) / Arizona State University (Publisher)
Created2023
Description
Abortion is a controversial topic internationally. Most current debates about abortion concern when, if at all, it should be legal. However, researchers have shown many times that after an abortion ban, maternal and infant mortalities rise significantly, as women who seek out abortions do so regardless of abortion legality. So, is it possible to reduce abortions in a population without delegalizing abortion and, if so, how? Why do some countries have higher abortion rates than others in the presence of the same law?This dissertation answers both questions. First, I present historical evidence in the first comprehensive comparative analysis of all 15 post-Soviet countries, which have very similar abortion laws originating from the Union of Soviet Socialist Republics (USSR). Second, I use those findings to build the first agent-based model (ABM) of unintended pregnancies in a hypothetical artificial population.
USSR was the only country in the world to complete its demographic transition through abortion instead of modern contraception, and the Soviet government passed the first law in the world to allow abortion upon request in 1920. After the USSR dissolution in 1991, post-Soviet countries maintained very similar abortion laws, but had very different abortion rates for most years. Analysis of fertility data from post-Soviet countries shows that the prevalence of some specific contraceptive methods, namely the rhythm method (r = 0.82), oral pill (r = 0.56), and male condom (r = 0.51) are most strongly correlated with high abortion rates, and that sex education is a factor that reduces the rates in otherwise similar countries (p = 0.02).
The ABM shows that even basic sex education results in fewer abortions than no sex education or abstinence-based sex education (p < 0.01). In scenarios without sex education, basic quality of post-abortion contraceptive counseling (PACC) is better than no PACC or low-quality PACC at reducing abortions (p < 0.01). Still, the higher the quality of sex education or PACC, the fewer abortions in the artificial population. The ABM is adaptive and policy makers can use it as a decision-support tool to make evidence-based policy decisions regarding abortion, and, potentially, other sociobiological phenomena with some adjustments to the code.
ContributorsZiganshina Lienhard, Dina A. (Author) / Maienschein, Jane (Thesis advisor) / Gaughan, Monica (Thesis advisor) / Laubichler, Manfred (Committee member) / Ellison, Karin (Committee member) / Arizona State University (Publisher)
Created2023
Description
Graph-structured data, ranging from social networks to financial transaction networks, from citation networks to gene regulatory networks, have been widely used for modeling a myriad of real-world systems. As a prevailing model architecture to model graph-structured data, graph neural networks (GNNs) has drawn much attention in both academic and industrial communities in the past decades. Despite their success in different graph learning tasks, existing methods usually rely on learning from ``big'' data, requiring a large amount of labeled data for model training. However, it is common that real-world graphs are associated with ``small'' labeled data as data annotation and labeling on graphs is always time and resource-consuming. Therefore, it is imperative to investigate graph machine learning (Graph ML) with low-cost human supervision for low-resource settings where limited or even no labeled data is available. This dissertation investigates a new research field -- Data-Efficient Graph Learning, which aims to push forward the performance boundary of graph machine learning (Graph ML) models with different kinds of low-cost supervision signals. To achieve this goal, a series of studies are conducted for solving different data-efficient graph learning problems, including graph few-shot learning, graph weakly-supervised learning, and graph self-supervised learning.
ContributorsDing, Kaize (Author) / Liu, Huan (Thesis advisor) / Xue, Guoliang (Committee member) / Yang, Yezhou (Committee member) / Caverlee, James (Committee member) / Arizona State University (Publisher)
Created2023
Description
This paper introduces Zenith, a statically typed, functional programming language that compiles to Lua modules. The goal of Zenith is to be used in tandem with Lua, as a secondary language, in which Lua developers can transition potentially unsound programs into Zenith instead. Here developers will be ensured a set of guarantees during compile time, which are provided through Zenith’s language design and type system. This paper formulates the reasoning behind the design choices in Zenith, based on prior work. This paper also provides a basic understanding and intuitions on the Hindley-Milner type system used in Zenith, and the functional programming data types used to encode unsound functions. With these ideas combined, the paper concludes on how Zenith can provide soundness and runtime safety as a language, and how Zenith may be used with Lua to create safe systems.
ContributorsShrestha, Abhash (Author) / De Luca, Gennaro (Thesis advisor) / Bansal, Ajay (Thesis advisor) / Chen, Yinong (Committee member) / Arizona State University (Publisher)
Created2023
Description
Large software tend to have a large number of configuration options that can be tuned to a varying degree in order to run the software in a specific way. These configuration options cause a change in the execution of the software, and therefore affect the code coverage of the software. This gives rise to the problem of understanding how much a certain configuration change affects the code coverage of the software in a measurable way. It also raises the question of effectively mapping code coverage to a configuration change. Solutions to these problems could give way to increasing efficiency in various areas of software security, like maximizing code coverage in fuzz testing and vulnerability identification in specific configurations.In this work, I perform analyze widely used software, such as the database cache `Redis' and web servers like `Nginx' and `Apache httpd'. I perform fuzz tests on multiple configurations of each of these software to measure the difference in code coverage caused by each configuration. I use Coverage Instrumentation to obtain traces for each software in their configurations, and then I analyze these traces to understand the configuration's impact on the software's code coverage.
In conclusion, I describe a method to measure how much code coverage differs for each configuration with respect to the default configuration of the software, and how certain configurations have a much larger difference in code coverage with respect to the default configuration than others, analyze the overlap in code coverage between the configurations and finally find the root causes of the differing code coverage.
ContributorsKumbhar, Swapnil (Author) / Shoshitaishvili, Yan (Thesis advisor) / Wang, Ruoyu (Committee member) / Xiao, Xusheng (Committee member) / Arizona State University (Publisher)
Created2023
Description
The prevalence of Autism Spectrum Disorders (ASD) has resulted in research on treatment efficacy, lending itself to quantitative analyses. In contrast, ethnographic studies have offered the benefits of analyzing lived experiences and exploring the uniqueness of ASD individuals. Using the Linguistic Ethnography (LE) perspective, this case study investigates the connection between multilingual practices, behavior management, and behavior-logging app usage. It looks at one cross-cultural family (Indonesian-Belgian) with two children diagnosed with autism. Multimodal data were collected for ten weeks virtually and five weeks in the family’s home in Belgium. The data collection focused on the family's multilingual and behavior management practices and specifically on mobile app usage to document the behaviors of the non-verbal son. Analytical frameworks were drawn from Multimodal Ethnography (Dicks et al., 2006) and Multimodal Discourse Analysis (Kress, 2011a). The findings indicated that linguistic and cultural diversity, both internally and externally, caused a layer of complexity in attending to a non-verbal child. The case study showed positive outcomes of multilingualism. However, it highlighted the struggle of building consistent communication between family members and health professionals, which affected the effort to find a successful treatment plan. The behavior logging app helped parents identify parts of the son’s behaviors and reflect on their behavior management strategies. However, it also underscored the real-life challenges of documentation. In this case, mobile technology may be more useful when supported by health professionals. Although the case study notes family successes, it calls attention to the extraordinary realities of cross-cultural ASD families that need more representation through ethnographic research.
ContributorsQodri, Asri Nurul (Author) / Smith, David Bryan (Thesis advisor) / SturtzSreetharan, Cindi (Committee member) / Dixon, Maria (Committee member) / Arizona State University (Publisher)
Created2023
Description
Code Generation is a task that has gained rapid progress in Natural Language Processing (NLP) research. This thesis focuses on the text-to-Structured Query Language (SQL) task, where the input is a question about a specific database and the output is the SQL that when executed will return the desired answer. The data creation process bottlenecks current text-to-SQL datasets. The technical knowledge required to understand and create SQL makes crowd-sourcing a dataset expensive and time-consuming. Thus, existing datasets do not provide a robust enough training set for state-of-the-art semantic parsing models. This thesis outlines my technique for generating a text-to-SQL dataset using GPT3 and prompt engineering techniques. My approach entails providing the Generative Pretrained Transformer 3 model (GPT-3) with particular instructions to build a rigorous text-to-SQL dataset. In this paper, I show that the created pairs have excellent quality and diversity, and when utilized as training data, they can enhance the accuracy of SQL generation models. I expect that my method will be of interest to academics in the disciplines of NLP because it can considerably reduce the time, effort, and cost necessary to produce large, high-quality text-to-SQL datasets. Furthermore, my approach can be extended to other tasks and domains to alleviate the burden of curating human-annotated data.
ContributorsKuznia, Kirby Charles (Author) / Baral, Chitta (Thesis advisor) / Blanco, Eduardo (Committee member) / Gopalan, Nakul (Committee member) / Arizona State University (Publisher)
Created2023
Description
Water, energy, and food are essential resources to sustain the development of the society. The Food-Energy-Water Nexus (FEW-Nexus) must account for synergies and trade-offs among these resources. The nexus concept highlights the importance of integrative solutions that secure supplies to meet demands sustainably. The existing frameworks and tools do not focus on formal model composability, a key capability for creating simulations created from separately developed models. The Knowledge Interchange Broker (KIB) approach is used to model the interactions among models to achieve composition flexibility for the FEW-Nexus.Domain experts generally use the Water Evaluation and Planning (WEAP) and Low Emissions Analysis Platform (LEAP) systems to study water and energy systems, respectively. The food part of FEW systems can be modeled inside the WEAP system. An internal linkage mechanism is available for combining and simulating WEAP and LEAP models. This mechanism is used for the validation and performance evaluation of independent modeling and simulation proposed in this research. The Componentized WEAP and LEAP RESTful frameworks are component-based representations for the legacy and closed-source WEAP and LEAP systems. These modularized systems simplify their use with other simulation frameworks.
This research proposes two interaction model frameworks based on the Knowledge Interchange Broker approach. First, an Algorithmic Interaction Model (Algorithmic-IM) was developed to integrate the WEAP and LEAP models. The Algorithmic-IM model can be defined via programming language and has a fixed cyclic execution protocol. However, this approach has tightly interwoven the interaction model with its execution and has limited support for flexibly creating model hierarchies. To overcome these restrictions, the system-theoretic Parallel DEVS formalism is used to develop a DEVS-Based Interaction Model (DEVS-IM). As in the Algorithmic-IM, the DEVS-IM is implemented as a RESTful framework, uses MongoDB for defining structural DEVS models, and supports automatic code generation for the DEVSSuite
simulator. The DEVS-IM offers modular, hierarchical structural modeling, reusability, flexibility, and maintainability for integrating disparate systems.
The Phoenix Active Management Area (AMA) is used to demonstrate the real-world application of the proposed research. Furthermore, the correctness and performance of the presented frameworks in this research are evaluated using the Phoenix-AMA model.
ContributorsFard, Mostafa D (Author) / Sarjoughian, Hessam S (Thesis advisor) / Barton, Michael (Committee member) / Sen, Arunabha (Committee member) / Zhao, Ming (Committee member) / Arizona State University (Publisher)
Created2023
Description
This dissertation presents novel solutions for improving the generalization capabilities of deep learning based computer vision models. Neural networks are known to suffer a large drop in performance when tested on samples from a different distribution than the one on which they were trained. The proposed solutions, based on latent space geometry and meta-learning, address this issue by improving the robustness of these models to distribution shifts. Through the use of geometrical alignment, state-of-the-art domain adaptation and source-free test-time adaptation strategies are developed. Additionally, geometrical alignment can allow classifiers to be progressively adapted to new, unseen test domains without requiring retraining of the feature extractors. The dissertation also presents algorithms for enabling in-the-wild generalization without needing access to any samples from the target domain. Other causes of poor generalization, such as data scarcity in critical applications and training data with high levels of noise and variance, are also explored. To address data scarcity in fine-grained computer vision tasks such as object detection, novel context-aware augmentations are suggested. While the first four chapters focus on general-purpose computer vision models, strategies are also developed to improve robustness in specific applications. The efficiency of training autonomous agents for visual navigation is improved by incorporating semantic knowledge, and the integration of domain experts' knowledge allows for the realization of a low-cost, minimally invasive generalizable automated rehabilitation system. Lastly, new tools for explainability and model introspection using counter-factual explainers trained through interval-based uncertainty calibration objectives are presented.
ContributorsThopalli, Kowshik (Author) / Turaga, Pavan (Thesis advisor) / Thiagarajan, Jayaraman J (Committee member) / Li, Baoxin (Committee member) / Yang, Yezhou (Committee member) / Arizona State University (Publisher)
Created2023
Description
The past decade witnessed the success of deep learning models in various applications of computer vision and natural language processing. This success can be predominantly attributed to the (i) availability of large amounts of training data; (ii) access of domain aware knowledge; (iii) i.i.d assumption between the train and target distributions and (iv) belief on existing metrics as reliable indicators of performance. When any of these assumptions are violated, the models exhibit brittleness producing adversely varied behavior. This dissertation focuses on methods for accurate model design and characterization that enhance process reliability when certain assumptions are not met. With the need to safely adopt artificial intelligence tools in practice, it is vital to build reliable failure detectors that indicate regimes where the model must not be invoked. To that end, an error predictor trained with a self-calibration objective is developed to estimate loss consistent with the underlying model. The properties of the error predictor are described and their utility in supporting introspection via feature importances and counterfactual explanations is elucidated. While such an approach can signal data regime changes, it is critical to calibrate models using regimes of inlier (training) and outlier data to prevent under- and over-generalization in models i.e., incorrectly identifying inliers as outliers and vice-versa. By identifying the space for specifying inliers and outliers, an anomaly detector that can effectively flag data of varying semantic complexities in medical imaging is next developed. Uncertainty quantification in deep learning models involves identifying sources of failure and characterizing model confidence to enable actionability. A training strategy is developed that allows the accurate estimation of model uncertainties and its benefits are demonstrated for active learning and generalization gap prediction. This helps identify insufficiently sampled regimes and representation insufficiency in models. In addition, the task of deep inversion under data scarce scenarios is considered, which in practice requires a prior to control the optimization. By identifying limitations in existing work, data priors powered by generative models and deep model priors are designed for audio restoration. With relevant empirical studies on a variety of benchmarks, the need for such design strategies is demonstrated.
ContributorsNarayanaswamy, Vivek Sivaraman (Author) / Spanias, Andreas (Thesis advisor) / J. Thiagarajan, Jayaraman (Committee member) / Berisha, Visar (Committee member) / Tepedelenlioğlu, Cihan (Committee member) / Arizona State University (Publisher)
Created2023