Computer Sciences Commons^™

Exact Models, Heuristics, And Supervised Learning Approaches For Vehicle Routing Problems, Zefeng Lyu

Doctoral Dissertations

This dissertation presents contributions to the field of vehicle routing problems by utilizing exact methods, heuristic approaches, and the integration of machine learning with traditional algorithms. The research is organized into three main chapters, each dedicated to a specific routing problem and a unique methodology. The first chapter addresses the Pickup and Delivery Problem with Transshipments and Time Windows, a variant that permits product transfers between vehicles to enhance logistics flexibility and reduce costs. To solve this problem, we propose an efficient mixed-integer linear programming model that has been shown to outperform existing ones. The second chapter discusses a practical …

Towards Robust Long-Form Text Generation Systems, Kalpesh Krishna

Doctoral Dissertations

Text generation is an important emerging AI technology that has seen significant research advances in recent years. Due to its closeness to how humans communicate, mastering text generation technology can unlock several important applications such as intelligent chat-bots, creative writing assistance, or newer applications like task-agnostic few-shot learning. Most recently, the rapid scaling of large language models (LLMs) has resulted in systems like ChatGPT, capable of generating fluent, coherent and human-like text. However, despite their remarkable capabilities, LLMs still suffer from several limitations, particularly when generating long-form text. In particular, (1) long-form generated text is filled with factual inconsistencies to …

Quantifying And Enhancing The Security Of Federated Learning, Virat Vishnu Shejwalkar

Doctoral Dissertations

Federated learning is an emerging distributed learning paradigm that allows multiple users to collaboratively train a joint machine learning model without having to share their private data with any third party. Due to many of its attractive properties, federated learning has received significant attention from academia as well as industry and now powers major applications, e.g., Google's Gboard and Assistant, Apple's Siri, Owkin's health diagnostics, etc. However, federated learning is yet to see widespread adoption due to a number of challenges. One such challenge is its susceptibility to poisoning by malicious users who aim to manipulate the joint machine learning …

Incremental Non-Greedy Clustering At Scale, Nicholas Monath

Doctoral Dissertations

Clustering is the task of organizing data into meaningful groups. Modern clustering applications such as entity resolution put several demands on clustering algorithms: (1) scalability to massive numbers of points as well as clusters, (2) incremental additions of data, (3) support for any user-specified similarity functions. Hierarchical clusterings are often desired as they represent multiple alternative flat clusterings (e.g., at different granularity levels). These tree-structured clusterings provide for both fine-grained clusters as well as uncertainty in the presence of newly arriving data. Previous work on hierarchical clustering does not fully address all three of the aforementioned desiderata. Work on incremental …

3d Shape Understanding And Generation, Matheus Gadelha

Doctoral Dissertations

In recent years, Machine Learning techniques have revolutionized solutions to longstanding image-based problems, like image classification, generation, semantic segmentation, object detection and many others. However, if we want to be able to build agents that can successfully interact with the real world, those techniques need to be capable of reasoning about the world as it truly is: a tridimensional space. There are two main challenges while handling 3D information in machine learning models. First, it is not clear what is the best 3D representation. For images, convolutional neural networks (CNNs) operating on raster images yield the best results in virtually …

Utilizing Graph Structure For Machine Learning, Stefan Dernbach

Doctoral Dissertations

The information age has led to an explosion in the size and availability of data. This data often exhibits graph-structure that is either explicitly defined, as in the web of a social network, or is implicitly defined and can be determined by measuring similarity between objects. Utilizing this graph-structure allows for the design of machine learning algorithms that reflect not only the attributes of individual objects but their relationships to every other object in the domain as well. This thesis investigates three machine learning problems and proposes novel methods that leverage the graph-structure inherent in the tasks. Quantum walk neural …

Reasoning About User Feedback Under Identity Uncertainty In Knowledge Base Construction, Ariel Kobren

Doctoral Dissertations

Intelligent, automated systems that are intertwined with everyday life---such as Google Search and virtual assistants like Amazon’s Alexa or Apple’s Siri---are often powered in part by knowledge bases (KBs), i.e., structured data repositories of entities, their attributes, and the relationships among them. Despite a wealth of research focused on automated KB construction methods, KBs are inevitably imperfect, with errors stemming from various points in the construction pipeline. Making matters more challenging, new data is created daily and must be integrated with existing KBs so that they remain up-to-date. As the primary consumers of KBs, human users have tremendous potential to …

Learning Latent Characteristics Of Data And Models Using Item Response Theory, John P. Lalor

Doctoral Dissertations

A supervised machine learning model is trained with a large set of labeled training data, and evaluated on a smaller but still large set of test data. Especially with deep neural networks (DNNs), the complexity of the model requires that an extremely large data set is collected to prevent overfitting. It is often the case that these models do not take into account specific attributes of the training set examples, but instead treat each equally in the process of model training. This is due to the fact that it is difficult to model latent traits of individual examples at the …

Neural Models For Information Retrieval Without Labeled Data, Hamed Zamani

Doctoral Dissertations

Recent developments of machine learning models, and in particular deep neural networks, have yielded significant improvements on several computer vision, natural language processing, and speech recognition tasks. Progress with information retrieval (IR) tasks has been slower, however, due to the lack of large-scale training data as well as neural network models specifically designed for effective information retrieval. In this dissertation, we address these two issues by introducing task-specific neural network architectures for a set of IR tasks and proposing novel unsupervised or \emph{weakly supervised} solutions for training the models. The proposed learning solutions do not require labeled training data. Instead, …

Extracting And Representing Entities, Types, And Relations, Patrick Verga

Doctoral Dissertations

Making complex decisions in areas like science, government policy, finance, and clinical treatments all require integrating and reasoning over disparate data sources. While some decisions can be made from a single source of information, others require considering multiple pieces of evidence and how they relate to one another. Knowledge graphs (KGs) provide a natural approach for addressing this type of problem: they can serve as long-term stores of abstracted knowledge organized around concepts and their relationships, and can be populated from heterogeneous sources including databases and text. KGs can facilitate higher level reasoning, influence the interpretation of new data, and …

From Optimization To Equilibration: Understanding An Emerging Paradigm In Artificial Intelligence And Machine Learning, Ian Gemp

Doctoral Dissertations

Many existing machine learning (ML) algorithms cannot be viewed as gradient descent on some single objective. The solution trajectories taken by these algorithms naturally exhibit rotation, sometimes forming cycles, a behavior that is not expected with (full-batch) gradient descent. However, these algorithms can be viewed more generally as solving for the equilibrium of a game with possibly multiple competing objectives. Moreover, some recent ML models, specifically generative adversarial networks (GANs) and its variants, are now explicitly formulated as equilibrium problems. Equilibrium problems present challenges beyond those encountered in optimization such as limit-cycles and chaotic attractors and are able to abstract …

Learning With Aggregate Data, Tao Sun

Doctoral Dissertations

Various real-world applications involve directly dealing with aggregate data. In this work, we study Learning with Aggregate Data from several perspectives and try to address their combinatorial challenges. At first, we study the problem of learning in Collective Graphical Models (CGMs), where only noisy aggregate observations are available. Inference in CGMs is NP- hard and we proposed an approximate inference algorithm. By solving the inference problems, we are empowered to build large-scale bird migration models, and models for human mobility under the differential privacy setting. Secondly, we consider problems given bags of instances and bag-level aggregate supervisions. Specifically, we study …

Applications Of Machine Learning In Nuclear Imaging And Radiation Detection, Shaikat Mahmood Galib

Doctoral Dissertations

"The main focus of this work is to use machine learning and data mining techniques to address some challenging problems that arise from nuclear data. Specifically, two problem areas are discussed: nuclear imaging and radiation detection. The techniques to approach these problems are primarily based on a variant of Artificial Neural Network (ANN) called Convolutional Neural Network (CNN), which is one of the most popular forms of 'deep learning' technique.

The first problem is about interpreting and analyzing 3D medical radiation images automatically. A method is developed to identify and quantify deformable image registration (DIR) errors from lung CT scans …

Machine Learning Methods For Activity Detection In Wearable Sensor Data Streams, Roy Adams

Doctoral Dissertations

Wearable wireless sensors have the potential for transformative impact on the fields of health and behavioral science. Recent advances in wearable sensor technology have made it possible to simultaneously collect multiple streams of physiological and context data from individuals in natural environments; however, extracting reliable high-level inferences from these raw data streams remains a key data analysis challenge. In this dissertation, we address three challenges that arise when trying to perform activity detection from wearable sensor streams. First, we address the challenge of learning from small amounts of noisy data by proposing a class of conditional random field models for …

Transfer Learning With Mixtures Of Manifolds, Thomas Boucher

Doctoral Dissertations

Advances in scientific instrumentation technology have increased the speed of data acquisition and the precision of sampling, creating an abundance of high-dimensional data sets. The ability to combine these disparate data sets and to transfer information between them is critical to accurate scientific analysis. Many modern-day instruments can record data at many thousands of channels, far greater than the actual degrees of freedom in the sample data. This makes manifold learning, a class of methods that exploit the observation that high-dimensional data tend to lie on lower-dimensional manifolds, especially well-suited to this transfer learning task. Existing manifold-based transfer learning methods …

Using Latent Variable Models To Improve Causal Estimation, Huseyin Oktay

Doctoral Dissertations

Estimating the causal effect of a treatment from data has been a key goal for a large number of studies in many domains. Traditionally, researchers use carefully designed randomized experiments for causal inference. However, such experiments can not only be costly in terms of time and money but also infeasible for some causal questions. To overcome these challenges, causal estimation methods from observational data have been developed by researchers from diverse disciplines and increasingly studies using such methods account for a large share in empirical work. Such growing interest has also brought together two arguably separate fields: machine learning and …

Deep-Learned Generative Representations Of 3d Shape Families, Haibin Huang

Doctoral Dissertations

Digital representations of 3D shapes are becoming increasingly useful in several emerging applications, such as 3D printing, virtual reality and augmented reality. However, traditional modeling softwares require users to have extensive modeling experience, artistic skills and training to handle their complex interfaces and perform the necessary low-level geometric manipulation commands. Thus, there is an emerging need for computer algorithms that help novice and casual users to quickly and easily generate 3D content. In this work, I will present deep learning algorithms that are capable of automatically inferring parametric representations of shape families, which can be used to generate new 3D …

Deep Energy-Based Models For Structured Prediction, David Belanger

Doctoral Dissertations

We introduce structured prediction energy networks (SPENs), a flexible frame- work for structured prediction. A deep architecture is used to define an energy func- tion over candidate outputs and predictions are produced by gradient-based energy minimization. This deep energy captures dependencies between labels that would lead to intractable graphical models, and allows us to automatically discover discrim- inative features of the structured output. Furthermore, practitioners can explore a wide variety of energy function architectures without having to hand-design predic- tion and learning methods for each model. This is because all of our prediction and learning methods interact with the energy …

Problems In Graph-Structured Modeling And Learning, James Atwood

Doctoral Dissertations

This thesis investigates three problems in graph-structured modeling and learning. We first present a method for efficiently generating large instances from nonlinear preferential attachment models of network structure. This is followed by a description of diffusion-convolutional neural networks, a new model for graph-structured data which is able to outperform probabilistic relational models and kernel-on-graph methods at node classification tasks. We conclude with an optimal privacy-protection method for users of online services that remains effective when users have poor knowledge of an adversary's behavior.

Learning From Pairwise Proximity Data, Hamid Dadkhahi

Doctoral Dissertations

In many areas of machine learning, the characterization of the input data is given by a form of proximity measure between data points. Examples of such representations are pairwise differences, pairwise distances, and pairwise comparisons. In this work, we investigate different learning problems on data represented in terms of such pairwise proximities. More specifically, we consider three problems: masking (feature selection) for dimensionality reduction, extension of the dimensionality reduction for time series, and online collaborative filtering. For each of these problems, we start with a form of pairwise proximity which is relevant in the problem at hand. We evaluate the …

Algorithms For First-Order Sparse Reinforcement Learning, Bo Liu

Doctoral Dissertations

This thesis presents a general framework for first-order temporal difference learning algorithms with an in-depth theoretical analysis. The main contribution of the thesis is the development and design of a family of first-order regularized temporal-difference (TD) algorithms using stochastic approximation and stochastic optimization. To scale up TD algorithms to large-scale problems, we use first-order optimization to explore regularized TD methods using linear value function approximation. Previous regularized TD methods often use matrix inversion, which requires cubic time and quadratic memory complexity. We propose two algorithms, sparse-Q and RO-TD, for on-policy and off-policy learning, respectively. These two algorithms exhibit linear computational …

Neuroscience-Inspired Dynamic Architectures, Catherine Dorothy Schuman

Doctoral Dissertations

Biological brains are some of the most powerful computational devices on Earth. Computer scientists have long drawn inspiration from neuroscience to produce computational tools. This work introduces neuroscience-inspired dynamic architectures (NIDA), spiking neural networks embedded in a geometric space that exhibit dynamic behavior. A neuromorphic hardware implementation based on NIDA networks, Dynamic Adaptive Neural Network Array (DANNA), is discussed. Neuromorphic implementations are one alternative/complement to traditional von Neumann computation. A method for designing/training NIDA networks, based on evolutionary optimization, is introduced. We demonstrate the utility of NIDA networks on classification tasks, a control task, and an anomaly detection task. There …

Epistemological Databases For Probabilistic Knowledge Base Construction, Michael Louis Wick

Doctoral Dissertations

Knowledge bases (KB) facilitate real world decision making by providing access to structured relational information that enables pattern discovery and semantic queries. Although there is a large amount of data available for populating a KB; the data must first be gathered and assembled. Traditionally, this integration is performed automatically by storing the output of an information extraction pipeline directly into a database as if this prediction were the ``truth.'' However, the resulting KB is often not reliable because (a) errors accumulate in the integration pipeline, and (b) they persist in the KB even after new information arrives that could rectify …

Learning With Joint Inference And Latent Linguistic Structure In Graphical Models, Jason Narad

Doctoral Dissertations

Constructing end-to-end NLP systems requires the processing of many types of linguistic information prior to solving the desired end task. A common approach to this problem is to construct a pipeline, one component for each task, with each system's output becoming input for the next. This approach poses two problems. First, errors propagate, and, much like the childhood game of "telephone", combining systems in this manner can lead to unintelligible outcomes. Second, each component task requires annotated training data to act as supervision for training the model. These annotations are often expensive and time-consuming to produce, may differ from each …

Causal Discovery For Relational Domains: Representation, Reasoning, And Learning, Marc Maier

Doctoral Dissertations

Many domains are currently experiencing the growing trend to record and analyze massive, observational data sets with increasing complexity. A commonly made claim is that these data sets hold potential to transform their corresponding domains by providing previously unknown or unexpected explanations and enabling informed decision-making. However, only knowledge of the underlying causal generative process, as opposed to knowledge of associational patterns, can support such tasks. Most methods for traditional causal discovery—the development of algorithms that learn causal structure from observational data—are restricted to representations that require limiting assumptions on the form of the data. Causal discovery has almost exclusively …

Adaptive Step-Sizes For Reinforcement Learning, William C. Dabney

Doctoral Dissertations

The central theme motivating this dissertation is the desire to develop reinforcement learning algorithms that “just work” regardless of the domain in which they are applied. The largest impediment to this goal is the sensitivity of reinforcement learning algorithms to the step-size parameter used to rescale incremental updates. Adaptive step-size algorithms attempt to reduce this sensitivity or eliminate the step-size parameter entirely by automatically adjusting the step size throughout the learning process. Such algorithms provide an alternative to the standard “guess-and-check” methods used to find parameters known as parameter tuning. However, the problems with parameter tuning are currently masked by …

Scaling Mcmc Inference And Belief Propagation To Large, Dense Graphical Models, Sameer Singh

Doctoral Dissertations

With the physical constraints of semiconductor-based electronics becoming increasingly limiting in the past decade, single-core CPUs have given way to multi-core and distributed computing platforms. At the same time, access to large data collections is progressively becoming commonplace due to the lowering cost of storage and bandwidth. Traditional machine learning paradigms that have been designed to operate sequentially on single processor architectures seem destined to become obsolete in this world of multi-core, multi-node systems and massive data sets. Inference for graphical models is one such example for which most existing algorithms are sequential in nature and are difficult to scale …

Incorporating Boltzmann Machine Priors For Semantic Labeling In Images And Videos, Andrew Kae

Doctoral Dissertations

Semantic labeling is the task of assigning category labels to regions in an image. For example, a scene may consist of regions corresponding to categories such as sky, water, and ground, or parts of a face such as eyes, nose, and mouth. Semantic labeling is an important mid-level vision task for grouping and organizing image regions into coherent parts. Labeling these regions allows us to better understand the scene itself as well as properties of the objects in the scene, such as their parts, location, and interaction within the scene. Typical approaches for this task include the conditional random field …

3d Robotic Sensing Of People: Human Perception, Representation And Activity Recognition, Hao Zhang

Doctoral Dissertations

The robots are coming. Their presence will eventually bridge the digital-physical divide and dramatically impact human life by taking over tasks where our current society has shortcomings (e.g., search and rescue, elderly care, and child education). Human-centered robotics (HCR) is a vision to address how robots can coexist with humans and help people live safer, simpler and more independent lives.

As humans, we have a remarkable ability to perceive the world around us, perceive people, and interpret their behaviors. Endowing robots with these critical capabilities in highly dynamic human social environments is a significant but very challenging problem in practical …

A Probabilistic Model Of Hierarchical Music Analysis, Phillip Benjamin Kirlin

Doctoral Dissertations

Schenkerian music theory supposes that Western tonal compositions can be viewed as hierarchies of musical objects. The process of Schenkerian analysis reveals this hierarchy by identifying connections between notes or chords of a composition that illustrate both the small- and large-scale construction of the music. We present a new probabilistic model of this variety of music analysis, details of how the parameters of the model can be learned from a corpus, an algorithm for deriving the most probable analysis for a given piece of music, and both quantitative and human-based evaluations of the algorithm's performance. In addition, we describe the …