Physical Sciences and Mathematics Commons^™

Detecting Anomalously Similar Entities In Unlabeled Data, Lisa D. Friedland

Doctoral Dissertations

In this work, the goal is to detect closely-linked entities within a data set. The entities of interest have a tie causing them to be similar, such as a shared origin or a channel of influence. Given a collection of people or other entities with their attributes or behavior, we identify unusually similar pairs, and we pose the question: Are these two people linked, or can their similarity be explained by chance? Computing similarities is a core operation in many domains, but two constraints differentiate our version of the problem. First, the score assigned to a pair should account for …

Stochastic Network Design: Models And Scalable Algorithms, Xiaojian Wu

Doctoral Dissertations

Many natural and social phenomena occur in networks. Examples include the spread of information, ideas, and opinions through a social network, the propagation of an infectious disease among people, and the spread of species within an interconnected habitat network. The ability to modify a phenomenon towards some desired outcomes has widely recognized benefits to our society and the economy. The outcome of a phenomenon is largely determined by the topology or properties of its underlying network. A decision maker can take management actions to modify a network and, therefore, change the outcome of the phenomenon. A management action is an …

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 …

Conditional Computation In Deep And Recurrent Neural Networks, Andrew Scott Davis

Doctoral Dissertations

Recently, deep learning models such as convolutional and recurrent neural networks have displaced state-of-the-art techniques in a variety of application domains. While the computationally heavy process of training is usually conducted on powerful graphics processing units (GPUs) distributed in large computing clusters, the resulting models can still be somewhat heavy, making deployment in resource- constrained environments potentially problematic. In this work, we build upon the idea of conditional computation, where the model is given the capability to learn how to avoid computing parts of the graph. This allows for models where the number of parameters (and in a sense, the …

Efficient Inference, Search And Evaluation For Latent Variable Models Of Text With Applications To Information Retrieval And Machine Translation, Kriste Krstovski

Doctoral Dissertations

Latent variable models of text, such as topic models, have been explored in many areas of natural language processing, information retrieval and machine translation to aid tasks such as exploratory data analysis, automated topic clustering and finding similar documents in mono- and multilingual collections. Many additional applications of these models, however, could be enabled by more efficient techniques for processing large datasets. In this thesis, we introduce novel methods that offer efficient inference, search and evaluation for latent variable models of text. We present efficient, online inference for representing documents in several languages in a common topic space and fast …

Extending Faceted Search To The Open-Domain Web, Weize Kong

Doctoral Dissertations

Faceted search enables users to navigate a multi-dimensional information space by combining keyword search with drill-down options in each facets. For example, when searching “computer monitor”' in an e-commerce site, users can select brands and monitor types from the the provided facets {“Samsung”, “Dell”, “Acer”, ...} and {“LET-Lit”, “LCD”, “OLED”, ...}. It has been used successfully for many vertical applications, including e-commerce and digital libraries. However, this idea is not well explored for general web search in an open-domain setting, even though it holds great potential for assisting multi-faceted queries and exploratory search. The goal of this work is to …

An Intelligent Robot And Augmented Reality Instruction System, Christopher M. Reardon

Doctoral Dissertations

Human-Centered Robotics (HCR) is a research area that focuses on how robots can empower people to live safer, simpler, and more independent lives. In this dissertation, I present a combination of two technologies to deliver human-centric solutions to an important population. The first nascent area that I investigate is the creation of an Intelligent Robot Instructor (IRI) as a learning and instruction tool for human pupils. The second technology is the use of augmented reality (AR) to create an Augmented Reality Instruction (ARI) system to provide instruction via a wearable interface.

To function in an intelligent and context-aware manner, both …

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 …