Artificial Intelligence and Robotics Commons^™

Attribution Robustness Of Neural Networks, Sunanda Gamage

Electronic Thesis and Dissertation Repository

While deep neural networks have demonstrated excellent learning capabilities, explainability of model predictions remains a challenge due to their black box nature. Attributions or feature significance methods are tools for explaining model predictions, facilitating model debugging, human-machine collaborative decision making, and establishing trust and compliance in critical applications. Recent work has shown that attributions of neural networks can be distorted by imperceptible adversarial input perturbations, which makes attributions unreliable as an explainability method. This thesis addresses the research problem of attribution robustness of neural networks and introduces novel techniques that enable robust training at scale.

Firstly, a novel generic framework …

A Target-Based And A Targetless Extrinsic Calibration Methods For Thermal Camera And 3d Lidar, Farhad Dalirani

Electronic Thesis and Dissertation Repository

This thesis introduces two novel methods for the extrinsic calibration of a thermal camera and a 3D LiDAR sensor, which are crucial for seamless data integration. The first method employs a distinctive calibration target, leveraging lines and plane equations correspondence in both modalities for a single pose, and incorporating more poses by matching the target's edges. It achieves reliable results, even with just one pose yielding 10.82% translation and 0.51-degree rotation errors. This outperforms alternative methods, which require eight pairs for similar results. The second method eliminates the need for a dedicated target. Instead, by collecting data during the sensor …

Learning Mortality Risk For Covid-19 Using Machine Learning And Statistical Methods, Shaoshi Zhang

Electronic Thesis and Dissertation Repository

This research investigates the mortality risk of COVID-19 patients across different variant waves, using the data from Centers for Disease Control and Prevention (CDC) websites. By analyzing the available data, including patient medical records, vaccination rates, and hospital capacities, we aim to discern patterns and factors associated with COVID-19-related deaths.

To explore features linked to COVID-19 mortality, we employ different techniques such as Filter, Wrapper, and Embedded methods for feature selection. Furthermore, we apply various machine learning methods, including support vector machines, decision trees, random forests, logistic regression, K-nearest neighbours, na¨ıve Bayes methods, and artificial neural networks, to uncover underlying …

Local Model Agnostic Xai Methodologies Applied To Breast Cancer Malignancy Predictions, Heather Hartley

Electronic Thesis and Dissertation Repository

This thesis examines current state-of-the-art Explainable Artificial Intelligence (XAI) methodologies applicable to breast cancer diagnostics, as well as local model-agnostic XAI methodologies more broadly. It is well known that AI is underutilized in healthcare due to the fact that black box AI methods are largely uninterpretable. The potential for AI to positively affect health care outcomes is massive, and AI adoption by medical practitioners and the community at large will translate to more desirable patient outcomes. The development of XAI is crucial to furthering the integration of AI within healthcare, as it will allow medical practitioners and regulatory bodies to …

Connectome-Constrained Artificial Neural Networks, Jacob Morra

Electronic Thesis and Dissertation Repository

In biological neural networks (BNNs), structure provides a set of guard rails by which function is constrained to solve tasks effectively, handle multiple stimuli simultaneously, adapt to noise and input variations, and preserve energy expenditure. Such features are desirable for artificial neural networks (ANNs), which are, unlike their organic counterparts, practically unbounded, and in many cases, initialized with random weights or arbitrary structural elements. In this dissertation, we consider an inductive base case for imposing BNN constraints onto ANNs. We select explicit connectome topologies from the fruit fly (one of the smallest BNNs) and impose these onto a multilayer perceptron …

Predicting Network Failures With Ai Techniques, Chandrika Saha

Electronic Thesis and Dissertation Repository

Network failure is the unintentional interruption of internet services, resulting in widespread client frustration. It is especially true for time-sensitive services in the healthcare industry, smart grid control, and mobility control, among others. In addition, the COVID-19 pandemic has compelled many businesses to operate remotely, making uninterrupted internet access essential. Moreover, Internet Service Providers (ISPs) lose millions of dollars annually due to network failure, which has a negative impact on their businesses. Currently, redundant network equipment is used as a restoration technique to resolve this issue of network failure. This technique requires a strategy for failure identification and prediction to …

Data-Driven Exploration Of Coarse-Grained Equations: Harnessing Machine Learning, Elham Kianiharchegani

Electronic Thesis and Dissertation Repository

In scientific research, understanding and modeling physical systems often involves working with complex equations called Partial Differential Equations (PDEs). These equations are essential for describing the relationships between variables and their derivatives, allowing us to analyze a wide range of phenomena, from fluid dynamics to quantum mechanics. Traditionally, the discovery of PDEs relied on mathematical derivations and expert knowledge. However, the advent of data-driven approaches and machine learning (ML) techniques has transformed this process. By harnessing ML techniques and data analysis methods, data-driven approaches have revolutionized the task of uncovering complex equations that describe physical systems. The primary goal in …

Weakly-Supervised Anomaly Detection In Surveillance Videos Based On Two-Stream I3d Convolution Network, Sareh Soltani Nejad

Electronic Thesis and Dissertation Repository

The widespread adoption of city surveillance systems has led to an increase in the use of surveillance videos for maintaining public safety and security. This thesis tackles the problem of detecting anomalous events in surveillance videos. The goal is to automatically identify abnormal events by learning from both normal and abnormal videos. Most of previous works consider any deviation from learned normal patterns as an anomaly, which may not always be valid since the same activity could be normal or abnormal under different circumstances. To address this issue, the thesis utilizes the Two-Stream Inflated 3D (I3D) Convolutional Networks to extract …

Towards Automated Mineral Identification In Martian Rocks From X-Ray Diffraction Patterns, Luke Tambakis

Electronic Thesis and Dissertation Repository

The CheMin (Chemistry and Mineralogy) instrument on the Curiosity rover has provided a rich set of X-ray diffraction (XRD) patterns from Martian rocks and regolith. These XRD patterns have allowed geologists to make exciting new discoveries about the mineralogy and the geological history of Mars. These discoveries pave the way for further Martian exploration and provide a deeper understanding of Martian geology. The Curiosity rover is very slow by design, travelling at about 4 cm/s. New, faster rovers are being developed to increase scientific throughput and exploration. XRD is valuable for future missions as it can produce new discov- eries …

Framework For Assessing Information System Security Posture Risks, Syed Waqas Hamdani

Electronic Thesis and Dissertation Repository

In today’s data-driven world, Information Systems, particularly the ones operating in regulated industries, require comprehensive security frameworks to protect against loss of confidentiality, integrity, or availability of data, whether due to malice, accident or otherwise. Once such a security framework is in place, an organization must constantly monitor and assess the overall compliance of its systems to detect and rectify any issues found. This thesis presents a technique and a supporting toolkit to first model dependencies between security policies (referred to as controls) and, second, devise models that associate risk with policy violations. Third, devise algorithms that propagate risk when …

An Approach To Lunar Regolith Particle Detection And Classification Using Deep Learning, Hira Nadeem

Electronic Thesis and Dissertation Repository

Lunar regolith, unconsolidated rock on the lunar surface, is made up of various particles. Understanding the quantities and locations of these particles on the lunar surface is of particular interest to planetary scientists for mission planning and science objectives. There is a limited supply of lunar regolith samples available on Earth for planetary scientists to characterize. Lunar rover missions over the next decade are expected to provide high-resolution images of the lunar surface. Deep learning can be leveraged to analyze lunar regolith from image data. An object detection model using transfer learning was developed to identify and classify particles of …

Investigating Improvements To Mesh Indexing, Anurag Bhattacharjee

Electronic Thesis and Dissertation Repository

The MEDLINE database currently comprises an extensive collection of over 35 million citations, with more than 1 million records being added each year [28]. The abundance of information available in the database presents a significant challenge in identifying and locating relevant research articles on a given search topic. This has prompted the development of various techniques and approaches aimed at improving the efficiency and effectiveness of information retrieval from the MEDLINE database. A search engine devoted to the research publications on MEDLINE is called PubMed. MeSH, or Medical Subject Headings, is a restricted vocabulary used by PubMed to categorize each …

Denoising-Based Domain Adaptation Network For Eeg Source Imaging, Runze Li

Electronic Thesis and Dissertation Repository

Electrophysiological source imaging (ESI) is a widespread and no-invasive technique in neuroscientific research and clinical diagnostics. It provides a well-established and high temporal resolution of source activity and gives the brain signal by analyzing the corresponding EEG signal.

However, it is still a major challenge to deal with the domain shift problem between the datasets of different subjects or sessions in ESI problem. Furthermore, the variable noise included in the EEG signals inevitably influence the accuracy of localization of source activity.

In this paper, we propose a novel denoising autoencoder-based unsupervised domain adaptation (DAE-UDA) algorithm to tackle these problems. To …

Citation Polarity Identification From Scientific Articles Using Deep Learning Methods, Souvik Kundu

Electronic Thesis and Dissertation Repository

The way in which research articles are cited reflects how previous work is utilized by other researchers or stakeholders and can indicate the impact of that work on subsequent experiments. Based on human intuition, citations can be perceived as positive, negative, or neutral. While current citation indexing systems provide information on the author and publication name of the cited article, as well as the citation count, they do not indicate the polarity of the citation. This study aims to identify the polarity of citations in scientific research articles using pre-trained language models like BERT, ELECTRA, RoBERTa, Bio-RoBERTa, SPECTER, ERNIE, LongFormer, …

Source-Free Domain Adaptation For Sleep Stage Classification, Yasmin Niknam

Electronic Thesis and Dissertation Repository

The popularity of machine learning algorithms has increased in recent years as data volumes have risen, algorithms have advanced, and computational power and storage have improved. EEG-based sleep staging has become one of the most active research areas over the last decade. Labeling each sleep stage manually is a labor-intensive and time-consuming process that requires expertise, making it susceptible to human error. In the meantime, training models on an unseen dataset remains challenging due to physiological differences between subjects and electrode sensor configurations. Unsupervised domain adaptation approaches may provide a solution to this problem by borrowing knowledge from a labeled …

Ai Applications On Planetary Rovers, Alexis David Pascual

Electronic Thesis and Dissertation Repository

The rise in the number of robotic missions to space is paving the way for the use of artificial intelligence and machine learning in the autonomy and augmentation of rover operations. For one, more rovers mean more images, and more images mean more data bandwidth required for downlinking as well as more mental bandwidth for analyzing the images. On the other hand, light-weight, low-powered microrover platforms are being developed to accommodate the drive for planetary exploration. As a result of the mass and power constraints, these microrover platforms will not carry typical navigational instruments like a stereocamera or a laser …

Reducing Negative Transfer Of Random Data In Source-Free Unsupervised Domain Adaptation, Anthony Wong

Electronic Thesis and Dissertation Repository

In domain adaptation, a model trained on one dataset (source domain) is applied to a different but related dataset (target domain). The most cutting-edge method is unsupervised source-free domain adaptation (SFDA), in which source data, source labels, and target labels are not available during adaptation. This thesis explores a realistic scenario where the target dataset includes some images that are unrelated to the adaptation process. This scenario can occur from errors in data collection or processing. We provide experiments and analysis to show that current state-of-the-art (SOTA) SFDA methods suffer significant performance drops under a specific domain adaptation setup when …

Attention-Based Multi-Source-Free Domain Adaptation For Eeg Emotion Recognition, Amir Hesam Salimnia

Electronic Thesis and Dissertation Repository

Electroencephalography (EEG) based emotion recognition in affective brain-computer interfaces has advanced significantly in recent years. Unsupervised domain adaptation (UDA) methods have been successfully used to mitigate the need for large amounts of training data, which is required due to the inter-subject variability of EEG signals. Typical UDA solutions require access to raw source data to leverage the knowledge learned from the labelled source domains (previous subjects) across the target domain (a new subject), raising privacy concerns. To tackle this issue, we propose Attention-based Multi-Source-Free Domain Adaptation (AMFDA) for EEG emotion recognition. AMFDA attempts to transfer knowledge of source models to …

Towards Parking Lot Occupancy Assessment Using Aerial Imagery And Computer Vision, John Jewell

Electronic Thesis and Dissertation Repository

Advances in Computer Vision and Aerial Imaging have enabled countless downstream applications. To this end, aerial imagery could be leveraged to analyze the usage of parking lots. This would enable retail centres to allocate space better and eliminate the parking oversupply problem. With this use case in mind, the proposed research introduces a novel framework for parking lot occupancy assessments. The framework consists of a pipeline of components that map a sequence of image sets spanning a parking lot at different time intervals to a parking lot turnover heatmap that encodes the frequency each parking stall was used. The pipeline …

Folk Theories, Recommender Systems, And Human-Centered Explainable Artificial Intelligence (Hcxai), Michael Ridley

Electronic Thesis and Dissertation Repository

This study uses folk theories to enhance human-centered “explainable AI” (HCXAI). The complexity and opacity of machine learning has compelled the need for explainability. Consumer services like Amazon, Facebook, TikTok, and Spotify have resulted in machine learning becoming ubiquitous in the everyday lives of the non-expert, lay public. The following research questions inform this study: What are the folk theories of users that explain how a recommender system works? Is there a relationship between the folk theories of users and the principles of HCXAI that would facilitate the development of more transparent and explainable recommender systems? Using the Spotify music …

Algorithmic Improvements In Deep Reinforcement Learning, Norman L. Tasfi

Electronic Thesis and Dissertation Repository

Reinforcement Learning (RL) has seen exponential performance improvements over the past decade, achieving super-human performance across many domains. Deep Reinforcement Learning (DRL), the combination of RL methods with deep neural networks (DNN) as function approximators, has unlocked much of this progress. The path to generalized artificial intelligence (GAI) will depend on deep learning (DL) and RL. However, much work is required before the technology reaches anything resembling GAI. Therefore, this thesis focuses on a subset of areas within RL that require additional research to advance the field, specifically: sample efficiency, planning, and task transfer. The first area, sample efficiency, refers …

Driver Behavior Analysis Based On Real On-Road Driving Data In The Design Of Advanced Driving Assistance Systems, Nima Khairdoost

Electronic Thesis and Dissertation Repository

The number of vehicles on the roads increases every day. According to the National Highway Traffic Safety Administration (NHTSA), the overwhelming majority of serious crashes (over 94 percent) are caused by human error. The broad aim of this research is to develop a driver behavior model using real on-road data in the design of Advanced Driving Assistance Systems (ADASs). For several decades, these systems have been a focus of many researchers and vehicle manufacturers in order to increase vehicle and road safety and assist drivers in different driving situations. Some studies have concentrated on drivers as the main actor in …

An Investigation Into Time Gazed At Traffic Objects By Drivers, Kolby R. Sarson

Electronic Thesis and Dissertation Repository

Several studies have considered driver’s attention for a multitude of distinct purposes, ranging from the analysis of a driver’s gaze and perception, to possible use in Advanced Driving Assistance Systems (ADAS). These works typically rely on simple definitions of what it means to “see,” considering a driver gazing upon an object for a single frame as being seen. In this work, we bolster this definition by introducing the concept of time. We consider a definition of ”seen” which requires an object to be gazed upon for a set length of time, or frames, before it can be considered as seen …

Potential Of Vision Transformers For Advanced Driver-Assistance Systems: An Evaluative Approach, Andrew Katoch

Electronic Thesis and Dissertation Repository

In this thesis, we examine the performance of Vision Transformers concerning the current state of Advanced Driving Assistance Systems (ADAS). We explore the Vision Transformer model and its variants on the problems of vehicle computer vision. Vision transformers show performance competitive to convolutional neural networks but require much more training data. Vision transformers are also more robust to image permutations than CNNs. Additionally, Vision Transformers have a lower pre-training compute cost but can overfit on smaller datasets more easily than CNNs. Thus we apply this knowledge to tune Vision transformers on ADAS image datasets, including general traffic objects, vehicles, traffic …

Understanding Deep Learning With Noisy Labels, Li Yi

Electronic Thesis and Dissertation Repository

Over the past decades, deep neural networks have achieved unprecedented success in image classification, which largely relies on the availability of correctly annotated large-scale datasets. However, collecting high-quality labels for large-scale datasets is expensive and time-consuming or even infeasible in practice. Approaches to addressing this issue include: acquiring labels from non-expert labelers, crowdsourcing-like platforms or other unreliable resources, where the label noise is inevitably involved. It becomes crucial to develop methods that are robust to label noise.

In this thesis, we study deep learning with noisy labels from two aspects. Specifically, the first part of this thesis, including two chapters, …

Towards The Development Of A Cost-Effective Image-Sensing-Smart-Parking Systems (Isensmap), Aakriti Sharma

Electronic Thesis and Dissertation Repository

Finding parking in a busy city has been a major daily problem in today’s busy life. Researchers have proposed various parking spot detection systems to overcome the problem of spending a long time searching for a parking spot. These works include a wide variety of sensors to detect the presence of a vehicle in a parking spot. These approaches are expensive to implement and ineffective in extreme weather conditions in an outdoor parking environment. As a result, a cost-effective, dependable, and time-saving parking solution is much more desirable. In this thesis, we proposed and developed an image processing-based real-time parking-spot …

Respiratory Pattern Analysis For Covid-19 Digital Screening Using Ai Techniques, Annita Tahsin Priyoti

Electronic Thesis and Dissertation Repository

Corona Virus (COVID-19) is a highly contagious respiratory disease that the World Health Organization (WHO) has declared a worldwide epidemic. This virus has spread worldwide, affecting various countries until now, causing millions of deaths globally. To tackle this public health crisis, medical professionals and researchers are working relentlessly, applying different techniques and methods. In terms of diagnosis, respiratory sound has been recognized as an indicator of one’s health condition. Our work is based on cough sound analysis. This study has included an in-depth analysis of the diagnosis of COVID-19 based on human cough sound. Based on cough audio samples from …

Improving Deep Entity Resolution By Constraints, Soudeh Nilforoushan

Electronic Thesis and Dissertation Repository

Entity resolutions the problem of finding duplicate data in a dataset and resolving possible differences and inconsistencies. ER is a long-standing data management and information retrieval problem and a core data integration and cleaning task. There are diverse solutions for ER that apply rule-based techniques, pairwise binary classification, clustering, and probabilistic inference, among other techniques. Deep learning (DL) has been extensively used for ER and has shown competitive performance compared to conventional ER solutions. The state-of-the-art (SOTA) ER solutions using DL are based on pairwise comparison and binary classification. They transform pairs of records into a latent space that can …

Reputation-Based Trust Assessment Of Transacting Service Components, Konstantinos Tsiounis

Electronic Thesis and Dissertation Repository

As Service-Oriented Systems rely for their operation on many different, and most often, distributed software components, a key issue that emerges is how one component can trust the services offered by another component. Here, the concept of trust is considered in the context of reputation systems and is viewed as a meta-requirement, that is, the level of belief a service requestor has that a service provider will provide the service in a way that meets the requestor’s expectations. We refer to the service offering components as service providers (SPs) and the service requesting components as service clients (SCs).

In this …

Developing Artificial Intelligence And Machine Learning To Support Primary Care Research And Practice, Jacqueline K. Kueper

Electronic Thesis and Dissertation Repository

This thesis was motivated by the potential to use "everyday data", especially that collected in electronic health records (EHRs) as part of healthcare delivery, to improve primary care for clients facing complex clinical and/or social situations. Artificial intelligence (AI) techniques can identify patterns or make predictions with these data, producing information to learn about and inform care delivery. Our first objective was to understand and critique the body of literature on AI and primary care. This was achieved through a scoping review wherein we found the field was at an early stage of maturity, primarily focused on clinical decision support …