Computer Engineering Commons^™

Exploring Gender Bias In Semantic Representations For Occupational Classification In Nlp: Techniques And Mitigation Strategies, Joseph Michael O'Carroll

Dissertations

Gender bias in Natural Language Processing (NLP) models is a non-trivial problem that can perpetuate and amplify existing societal biases. This thesis investigates gender bias in occupation classification and explores the effectiveness of different debiasing methods for language models to reduce the impact of bias in the model’s representations. The study employs a data-driven empirical methodology focusing heavily on experimentation and result investigation. The study uses five distinct semantic representations and models with varying levels of complexity to classify the occupation of individuals based on their biographies.

Probability Expressions In Ai Decision Support: Impacts On Human+Ai Team Performance, Elias Spinn

Dissertations

AI decision support systems aim to assist people in highly complex and consequential domains to make efficient, effective, and high-quality decisions. AI alone cannot be guaranteed to be correct in these complex decision tasks, and a human is often needed to ensure decision accuracy. The ambition is for these human+ AI teams to perform better together than either would individually. To realise this, decision makers must trust their AI partners appropriately, knowing when to rely on their recommendations and when to be sceptical. However, research has shown that decision makers often either mistrust and underutilise these systems, or trust them …

Artificial Neural Networks And Their Applications To Intelligent Fault Diagnosis Of Power Transmission Lines, Fatemeh Mohammadi Shakiba

Dissertations

Over the past thirty years, the idea of computing based on models inspired by human brains and biological neural networks emerged. Artificial neural networks play an important role in the field of machine learning and hold the key to the success of performing many intelligent tasks by machines. They are used in various applications such as pattern recognition, data classification, stock market prediction, aerospace, weather forecasting, control systems, intelligent automation, robotics, and healthcare. Their architectures generally consist of an input layer, multiple hidden layers, and one output layer. They can be implemented on software or hardware. Nowadays, various structures with …

Local Learning Algorithms For Stochastic Spiking Neural Networks, Bleema Rosenfeld

Dissertations

This dissertation focuses on the development of machine learning algorithms for spiking neural networks, with an emphasis on local three-factor learning rules that are in keeping with the constraints imposed by current neuromorphic hardware. Spiking neural networks (SNNs) are an alternative to artificial neural networks (ANNs) that follow a similar graphical structure but use a processing paradigm more closely modeled after the biological brain in an effort to harness its low power processing capability. SNNs use an event based processing scheme which leads to significant power savings when implemented in dedicated neuromorphic hardware such as Intel’s Loihi chip.

This work …

Statistics-Based Anomaly Detection And Correction Method For Amazon Customer Reviews, Ishani Chatterjee

Dissertations

People nowadays use the Internet to project their assessments, impressions, ideas, and observations about various subjects or products on numerous social networking sites. These sites serve as a great source of gathering information for data analytics, sentiment analysis, natural language processing, etc. The most critical challenge is interpreting this data and capturing the sentiment behind these expressions. Sentiment analysis is analyzing, processing, concluding, and inferencing subjective texts with the views. Companies use sentiment analysis to understand public opinions, perform market research, analyze brand reputation, recognize customer experiences, and study social media influence. According to the different needs for aspect granularity, …

On Resource-Efficiency And Performance Optimization In Big Data Computing And Networking Using Machine Learning, Wuji Liu

Dissertations

Due to the rapid transition from traditional experiment-based approaches to large-scale, computational intensive simulations, next-generation scientific applications typically involve complex numerical modeling and extreme-scale simulations. Such model-based simulations oftentimes generate colossal amounts of data, which must be transferred over high-performance network (HPN) infrastructures to remote sites and analyzed against experimental or observation data on high-performance computing (HPC) facility. Optimizing the performance of both data transfer in HPN and simulation-based model development on HPC is critical to enabling and accelerating knowledge discovery and scientific innovation. However, such processes generally involve an enormous set of attributes including domain-specific model parameters, network transport …

Data-Driven Learning For Robot Physical Intelligence, Leidi Zhao

Dissertations

The physical intelligence, which emphasizes physical capabilities such as dexterous manipulation and dynamic mobility, is essential for robots to physically coexist with humans. Much research on robot physical intelligence has achieved success on hyper robot motor capabilities, but mostly through heavily case-specific engineering. Meanwhile, in terms of robot acquiring skills in a ubiquitous manner, robot learning from human demonstration (LfD) has achieved great progress, but still has limitations handling dynamic skills and compound actions. In this dissertation, a composite learning scheme which goes beyond LfD and integrates robot learning from human definition, demonstration, and evaluation is proposed. This method tackles …

Data Mining Of Unstructured Textual Information In Transportation Safety Domain: Exploring Methods, Opportunities And Limitations, Keneth Morgan Kwayu

Dissertations

The unprecedented increase in volume and influx of structured and unstructured data has overwhelmed conventional data management system capabilities in organizing, analyzing, and procuring useful information in a timely fashion. Structured data sources have a pre-defined pattern that makes data preprocessing and information retrieval tasks relatively easy for the current technologies that have been designed to handle structured and repeatable data. Unlike structured data, unstructured data usually exists in an unorganized format that offers no or little insight unless indexed and stored in an organized fashion. The inherent format of unstructured data exacerbates difficulties in data preprocessing and information extraction. …

Hybrid Deep Neural Networks For Mining Heterogeneous Data, Xiurui Hou

Dissertations

In the era of big data, the rapidly growing flood of data represents an immense opportunity. New computational methods are desired to fully leverage the potential that exists within massive structured and unstructured data. However, decision-makers are often confronted with multiple diverse heterogeneous data sources. The heterogeneity includes different data types, different granularities, and different dimensions, posing a fundamental challenge in many applications. This dissertation focuses on designing hybrid deep neural networks for modeling various kinds of data heterogeneity.

The first part of this dissertation concerns modeling diverse data types, the first kind of data heterogeneity. Specifically, image data and …

Live Media Production: Multicast Optimization And Visibility For Clos Fabric In Media Data Centers, Ammar Latif

Dissertations

Media production data centers are undergoing a major architectural shift to introduce digitization concepts to media creation and media processing workflows. Content companies such as NBC Universal, CBS/Viacom and Disney are modernizing their workflows to take advantage of the flexibility of IP and virtualization.

In these new environments, multicast is utilized to provide point-to-multi-point communications. In order to build point-to-multi-point trees, Multicast has an established set of control protocols such as IGMP and PIM. The existing multicast protocols do not optimize multicast tree formation for maximizing network throughput which lead to decreased fabric utilization and decreased total number of admitted …

Changing The Focus: Worker-Centric Optimization In Human-In-The-Loop Computations, Mohammadreza Esfandiari

Dissertations

A myriad of emerging applications from simple to complex ones involve human cognizance in the computation loop. Using the wisdom of human workers, researchers have solved a variety of problems, termed as “micro-tasks” such as, captcha recognition, sentiment analysis, image categorization, query processing, as well as “complex tasks” that are often collaborative, such as, classifying craters on planetary surfaces, discovering new galaxies (Galaxyzoo), performing text translation. The current view of “humans-in-the-loop” tends to see humans as machines, robots, or low-level agents used or exploited in the service of broader computation goals. This dissertation is developed to shift the focus back …

Analyzing Evolution Of Rare Events Through Social Media Data, Xiaoyu Lu

Dissertations

Recently, some researchers have attempted to find a relationship between the evolution of rare events and temporal-spatial patterns of social media activities. Their studies verify that the relationship exists in both time and spatial domains. However, few of those studies can accurately deduce a time point when social media activities are most highly affected by a rare event because producing an accurate temporal pattern of social media during the evolution of a rare event is very difficult. This work expands the current studies along three directions. Firstly, we focus on the intensity of information volume and propose an innovative clustering …

Analyzing Twitter Feeds To Facilitate Crises Informatics And Disaster Response During Mass Emergencies, Arshdeep Kaur

Dissertations

It is a common practice these days for general public to use various micro-blogging platforms, predominantly Twitter, to share ideas, opinions and information about things and life. Twitter is also being increasingly used as a popular source of information sharing during natural disasters and mass emergencies to update and communicate the extent of the geographic phenomena, report the affected population and casualties, request or provide volunteering services and to share the status of disaster recovery process initiated by humanitarian-aid and disaster-management organizations. Recent research in this area has affirmed the potential use of such social media data for various disaster …

From Business Understanding To Deployment: An Application Of Machine Learning Algorithms To Forecast Customer Visits Per Hour To A Fast-Casual Restaurant In Dublin, Odunayo David Adedeji

Dissertations

This research project identifies the significant factors that affects the number of customer visits to a fast-casual restaurant every hour and proceeds to develop several machine learning models to forecast customer visits. The core value proposition of fast-casual restaurants is quality food delivered at speed which means they have to prepare meals in advance of customers visit but the problem with this approach is in forecasting future demand, under estimating demand could lead to inadequate meal preparation which would leave customers unsatisfied while over estimation of demand could lead to wastage especially with restaurants having to comply with food safety …

Application Of Synthetic Informative Minority Over-Sampling (Simo) Algorithm Leveraging Support Vector Machine (Svm) On Small Datasets With Class Imbalance, Akshatha Fakkeriah Kallappanamatt

Dissertations

Developing predictive models for classification problems considering imbalanced datasets is one of the basic difficulties in data mining and decision-analytics. A classifier’s performance will decline dramatically when applied to an imbalanced dataset. Standard classifiers such as logistic regression, Support Vector Machine (SVM) are appropriate for balanced training sets whereas provides suboptimal classification results when used on unbalanced dataset. Performance metric with prediction accuracy encourages a bias towards the majority class, while the rare instances remain unknown though the model contributes a high overall precision. There are chances where minority instances might be treated as noise and vice versa. (Haixiang et …

Subspace Methods For Portfolio Design, Onur Yilmaz

Dissertations

Financial signal processing (FSP) is one of the emerging areas in the field of signal processing. It is comprised of mathematical finance and signal processing. Signal processing engineers consider speech, image, video, and price of a stock as signals of interest for the given application. The information that they will infer from raw data is different for each application. Financial engineers develop new solutions for financial problems using their knowledge base in signal processing. The goal of financial engineers is to process the harvested financial signal to get meaningful information for the purpose.

Designing investment portfolios have always been at …

An Understanding Of Student Satisfaction, Lorraine Sweeney

Dissertations

Retention is a challenge for all third level institutions and retention rates remain higher than colleges would like them to be, this has intensified in recent years as participants in higher education has increased and diversified. Third level institutions which would not only benefit from increased fees but also through low cost word of mouth promotion and an enhanced reputation. As such, an important concern for colleges is retaining students and understanding the reasons why students may choose to leave a program. While student satisfaction and retention is a well researched topic there remains questions to be answered in terms …

Eliciting Knowledge Bases With Defeasible Reasoning: A Comparative Analysis With Machine Learning, Peter Keogh

Dissertations

This thesis compares the ability of an implementation of Defeasible Reasoning (via Argumentation Theory) to model a construct (mental workload) with Machine Learning. In order to perform this comparison a defeasible reasoning system was designed and implemented in software. This software was used to elicit a knowledge base from an expert in an experiment which was then compared with machine learning. The central findings of this thesis were that the knowledge based approach was better at predicting an objective performance measure, time, than machine learning. However, machine learning was better equiped to identify another object measure task membership. The knowledge …

Sports Data Mining Technology Used In Basketball Outcome Prediction, Chenjie Cao

Dissertations

Driven by the increasing comprehensive data in sports datasets and data mining technique successfully used in different area, sports data mining technique emerges and enables us to find hidden knowledge to impact the sport industry. In many instances, predicting the outcomes of sporting events has always been a challenging and attractive work and is therefore drawing a wide concern to conduct research in this field. This project focuses on using machine learning algorithms to build a model for predicting the NBA game outcomes and the algorithms involve Simple Logistics Classifier, Artificial Neural Networks, SVM and Naïve Bayes. In order to …