Matching Items (911)
Filtering by
- All Subjects: Computer Science
- Creators: Arizona State University
- Resource Type: Text
Description
In image classification tasks, images are often corrupted by spatial transformationslike translations and rotations. In this work, I utilize an existing method that uses
the Fourier series expansion to generate a rotation and translation invariant representation of closed contours found in sketches, aiming to attenuate the effects of distribution shift caused by the aforementioned transformations. I use this technique to
transform input images into one of two different invariant representations, a Fourier
series representation and a corrected raster image representation, prior to passing
them to a neural network for classification. The architectures used include convolutional neutral networks (CNNs), multi-layer perceptrons (MLPs), and graph neural
networks (GNNs). I compare the performance of this method to using data augmentation during training, the standard approach for addressing distribution shift, to see
which strategy yields the best performance when evaluated against a test set with
rotations and translations applied. I include experiments where the augmentations
applied during training both do and do not accurately reflect the transformations encountered at test time. Additionally, I investigate the robustness of both approaches
to high-frequency noise. In each experiment, I also compare training efficiency across
models. I conduct experiments on three data sets, the MNIST handwritten digit
dataset, a custom dataset (QD-3) consisting of three classes of geometric figures from
the Quick, Draw! hand-drawn sketch dataset, and another custom dataset (QD-345)
featuring sketches from all 345 classes found in Quick, Draw!. On the smaller problem space of MNIST and QD-3, the networks utilizing the Fourier-based technique to
attenuate distribution shift perform competitively with the standard data augmentation strategy. On the more complex problem space of QD-345, the networks using the
Fourier technique do not achieve the same test performance as correctly-applied data
augmentation. However, they still outperform instances where train-time augmentations mis-predict test-time transformations, and outperform a naive baseline model
where no strategy is used to attenuate distribution shift. Overall, this work provides
evidence that strategies which attempt to directly mitigate distribution shift, rather
than simply increasing the diversity of the training data, can be successful when
certain conditions hold.
ContributorsWatson, Matthew (Author) / Yang, Yezhou YY (Thesis advisor) / Kerner, Hannah HK (Committee member) / Yang, Yingzhen YY (Committee member) / Arizona State University (Publisher)
Created2023
Description
Open Information Extraction (OIE) is a subset of Natural Language Processing (NLP) that constitutes the processing of natural language into structured and machine-readable data. This thesis uses data in Resource Description Framework (RDF) triple format that comprises of a subject, predicate, and object. The extraction of RDF triples from natural language is an essential step towards importing data into web ontologies as part of the linked open data cloud on the Semantic web. There have been a number of related techniques for extraction of triples from plain natural language text including but not limited to ClausIE, OLLIE, Reverb, and DeepEx. This proposed study aims to reduce the dependency on conventional machine learning models since they require training datasets, and the models are not easily customizable or explainable. By leveraging a context-free grammar (CFG) based model, this thesis aims to address some of these issues while minimizing the trade-offs on performance and accuracy. Furthermore, a deep-dive is conducted to analyze the strengths and limitations of the proposed approach.
ContributorsSingh, Varun (Author) / Bansal, Srividya (Thesis advisor) / Bansal, Ajay (Committee member) / Mehlhase, Alexandra (Committee member) / Arizona State University (Publisher)
Created2023
Description
Instruction tuning of language models has demonstrated the ability to enhance model generalization to unseen tasks via in-context learning using a few examples. However, typical supervised learning still requires a plethora of training data for downstream or “Held-in” tasks. Often in real-world situations, there is a scarcity of data available for finetuning, falling somewhere between few shot inference and fully supervised finetuning. In this work, I demonstrate the sample efficiency of instruction tuned models over various tasks by estimating the minimal training data required by downstream or “Held-In” tasks to perform transfer learning and match the performance of state-of-the-art (SOTA) supervised models. I conduct experiments on 119 tasks from Super Natural Instructions (SuperNI) in both the single task learning / Expert Modelling (STL) and multi task learning (MTL) settings. My findings reveal that, in the STL setting, instruction tuned models equipped with 25% of the downstream train data surpass the SOTA performance on the downstream tasks. In the MTL setting, an instruction tuned model trained on only 6% of downstream training data achieve SOTA, while using 100% of the training data results in a 3.69% points improvement (ROUGE-L 74.68) over the previous SOTA. I conduct an analysis on T5 vs Tk-Instruct by developing several baselines to demonstrate that instruction tuning aids in increasing both sample efficiency and transfer learning. Additionally, I observe a consistent ∼ 4% performance increase in both settings when pre-finetuning is performed with instructions. Finally, I conduct a categorical study and find that contrary to previous results, tasks in the question rewriting and title generation categories suffer from instruction tuning.
ContributorsGupta, Himanshu (Author) / Baral, Chitta Dr (Thesis advisor) / Mitra, Arindam Dr (Committee member) / Gopalan, Nakul Dr (Committee member) / Arizona State University (Publisher)
Created2023
Description
The evolution of technology, including the proliferation of the Internet of Things (IoT), advanced sensors, intelligent systems, and more, has paved the way for the establishment of smart homes. These homes bring a new era of automation with interconnected devices, offering increased services. However, they also introduce data security and device management challenges. Current smart home technologies are susceptible to security violations, leaving users vulnerable to data compromise, privacy invasions, and physical risks. These systems often fall short in implementing stringent data security safeguards, and the user control process is complex. In this thesis, an approach is presented to improve smart home security by integrating private blockchain technology with situational awareness access control. Using blockchain technology ensures transparency and immutability in data transactions. Transparency from the blockchain enables meticulous tracking of data access, modifications, and policy changes. The immutability of blockchain is utilized to strengthen the integrity of data, deterring, and preventing unauthorized alterations. While the designed solution leverages these specific blockchain features, it consciously does not employ blockchain's decentralization due to the limited computational resources of IoT devices and the focused requirement for centralized management within a smart home context. Additionally, situational awareness facilitates the dynamic adaptation of access policies.
The strategies in this thesis excel beyond existing solutions, providing fine-grained access control, reliable transaction data storage, data ownership, audibility, transparency, access policy, and immutability. This approach is thoroughly evaluated against existing smart home security improvement solutions.
ContributorsLin, Zhicheng (Author) / Yau, Stephen S. (Thesis advisor) / Baek, Jaejong (Committee member) / Ghayekhloo, Samira (Committee member) / Arizona State University (Publisher)
Created2023
Description
Pan Tilt Traffic Cameras (PTTC) are a vital component of traffic managementsystems for monitoring/surveillance. In a real world scenario, if a vehicle is in pursuit
of another vehicle or an accident has occurred at an intersection causing traffic
stoppages, accurate and venerable data from PTTC is necessary to quickly localize
the cars on a map for adept emergency response as more and more traffic systems are
getting automated using machine learning concepts. However, the position(orientation)
of the PTTC with respect to the environment is often unknown as most of them
lack Inertial Measurement Units or Encoders. Current State Of the Art systems 1.
Demand high performance compute and use carbon footprint heavy Deep Neural
Networks(DNN), 2. Are only applicable to scenarios with appropriate lane markings or
only roundabouts, 3. Demand complex mathematical computations to determine focal
length and optical center first before determining the pose. A compute light approach
"TIPANGLE" is presented in this work. The approach uses the concept of Siamese
Neural Networks(SNN) encompassing simple mathematical functions i.e., Euclidian
Distance and Contrastive Loss to achieve the objective. The effectiveness of the
approach is reckoned with a thorough comparison study with alternative approaches
and also by executing the approach on an embedded system i.e., Raspberry Pi 3.
ContributorsJagadeesha, Shreehari (Author) / Shrivastava, Aviral (Thesis advisor) / Gopalan, Nakul (Committee member) / Arora, Aman (Committee member) / Arizona State University (Publisher)
Created2023
Description
This dissertation is an examination of collective systems of computationally limited agents that require coordination to achieve complex ensemble behaviors or goals. The design of coordination strategies can be framed as multiagent optimization problems, which are addressed in this work from both theoretical and practical perspectives. The primary foci of this study are models where computation is distributed over the agents themselves, which are assumed to possess onboard computational capabilities. There exist many assumption variants for distributed models, including fairness and concurrency properties. In general, there is a fundamental trade-off whereby weakening model assumptions increases the applicability of proposed solutions, while also increasing the difficulty of proving theoretical guarantees. This dissertation aims to produce a deeper understanding of this trade-off with respect to multiagent optimization and scalability in distributed settings. This study considers four multiagent optimization problems. The model assumptions begin with fully centralized computation for the all-or-nothing multicommodity flow problem, then progress to synchronous distributed models through examination of the unmapped multivehicle routing problem and the distributed target localization problem. The final model is again distributed but assumes an unfair asynchronous adversary in the context of the energy distribution problem for programmable matter. For these problems, a variety of algorithms are presented, each of which is grounded in a theoretical foundation that permits formal guarantees regarding correctness, running time, and other critical properties. These guarantees are then validated with in silico simulations and (in some cases) physical experiments, demonstrating empirically that they may carry over to the real world. Hence, this dissertation bridges a portion of the predictability-practicality gap with respect to multiagent optimization problems.
ContributorsWeber, Jamison Wayne (Author) / Richa, Andréa W (Thesis advisor) / Bertsekas, Dimitri P (Committee member) / Murphey, Todd D (Committee member) / Jiang, Zilin (Committee member) / Arizona State University (Publisher)
Created2023
Description
Due to the internet being in its infancy, there is no consensus regarding policy approaches that various countries have taken. These policies range from strict government control to liberal access to the internet which makes protecting individual private data difficult. There are too many loopholes and various forms of policy on how to approach protecting data. There must be effort by both the individual, government, and private entities by using theoretical mixed methods to approach protecting oneself properly online.
ContributorsPeralta, Christina A (Author) / Scheall, Scott (Thesis advisor) / Hollinger, Keith (Thesis advisor) / Alozie, Nicholas (Committee member) / Arizona State University (Publisher)
Created2023
Description
Social media platforms provide a rich environment for analyzing user behavior. Recently, deep learning-based methods have been a mainstream approach for social media analysis models involving complex patterns. However, these methods are susceptible to biases in the training data, such as participation inequality. Basically, a mere 1% of users generate the majority of the content on social networking sites, while the remaining users, though engaged to varying degrees, tend to be less active in content creation and largely silent. These silent users consume and listen to information that is propagated on the platform.However, their voice, attitude, and interests are not reflected in the online content, making the decision of the current methods predisposed towards the opinion of the active users. So models can mistake the loudest users for the majority. To make the silent majority heard is to reveal the true landscape of the platform. In this dissertation, to compensate for this bias in the data, which is related to user-level data scarcity, I introduce three pieces of research work. Two of these proposed solutions deal with the data on hand while the other tries to augment the current data. Specifically, the first proposed approach modifies the weight of users' activity/interaction in the input space, while the second approach involves re-weighting the loss based on the users' activity levels during the downstream task training. Lastly, the third approach uses large language models (LLMs) and learns the user's writing behavior to expand the current data. In other words, by utilizing LLMs as a sophisticated knowledge base, this method aims to augment the silent user's data.
ContributorsKarami, Mansooreh (Author) / Liu, Huan (Thesis advisor) / Sen, Arunabha (Committee member) / Davulcu, Hasan (Committee member) / Mancenido, Michelle V. (Committee member) / Arizona State University (Publisher)
Created2023
Description
Human civilization within the last two decades has largely transformed into an online one, with many of its associated activities taking place on computers and complex networked systems -- their analog and real-world equivalents having been rendered obsolete.These activities run the gamut from the ordinary and mundane, like ordering food, to complex and large-scale, such as those involving critical infrastructure or global trade and communications.
Unfortunately, the activities of human civilization also involve criminal, adversarial, and malicious ones with the result that they also now have their digital equivalents.
Ransomware, malware, and targeted cyberattacks are a fact of life today and are instigated not only by organized criminal gangs, but adversarial nation-states and organizations as well.
Needless to say, such actions result in disastrous and harmful real-world consequences. As the complexity and variety of software has evolved, so too has the ingenuity of attacks that exploit them; for example modern cyberattacks typically involve sequential exploitation of multiple software vulnerabilities.Compared to a decade ago, modern software stacks on personal computers, laptops, servers, mobile phones, and even Internet of Things (IoT) devices involve a dizzying array of interdependent programs and software libraries, with each of these components presenting attractive attack-surfaces for adversarial actors.
However, the responses to this still rely on paradigms that can neither react quickly enough nor scale to increasingly dynamic, ever-changing, and complex software environments.
Better approaches are therefore needed, that can assess system readiness and vulnerabilities, identify potential attack vectors and strategies (including ways to counter them), and proactively detect vulnerabilities in complex software before they can be exploited. In this dissertation, I first present a mathematical model and associated algorithms to identify attacker strategies for sequential cyberattacks based on attacker state, attributes and publicly-available vulnerability information.Second, I extend the model and design algorithms to help identify defensive courses of action against attacker strategies.
Finally, I present my work to enhance the ability of coverage-based fuzzers to identify software vulnerabilities by providing visibility into complex, internal program-states.
ContributorsPaliath, Vivin Suresh (Author) / Doupe, Adam (Thesis advisor) / Shoshitaishvili, Yan (Thesis advisor) / Wang, Ruoyu (Committee member) / Shakarian, Paulo (Committee member) / Arizona State University (Publisher)
Created2023
Description
When looking at drawings of graphs, questions about graph density, community structures, local clustering and other graph properties may be of critical importance for analysis. While graph layout algorithms have focused on minimizing edge crossing, symmetry, and other such layout properties, there is not much known about how these algorithms relate to a user’s ability to perceive graph properties for a given graph layout. This study applies previously established methodologies for perceptual analysis to identify which graph drawing layout will help the user best perceive a particular graph property. A large scale (n = 588) crowdsourced experiment is conducted to investigate whether the perception of two graph properties (graph density and average local clustering coefficient) can be modeled using Weber’s law. Three graph layout algorithms from three representative classes (Force Directed - FD, Circular, and Multi-Dimensional Scaling - MDS) are studied, and the results of this experiment establish the precision of judgment for these graph layouts and properties. The findings demonstrate that the perception of graph density can be modeled with Weber’s law. Furthermore, the perception of the average clustering coefficient can be modeled as an inverse of Weber’s law, and the MDS layout showed a significantly different precision of judgment than the FD layout.
ContributorsSoni, Utkarsh (Author) / Maciejewski, Ross (Thesis advisor) / Kobourov, Stephen (Committee member) / Sefair, Jorge (Committee member) / Arizona State University (Publisher)
Created2018