Computer Science Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/2756
Browse
4 results
Search Results
Item Algorithms and Data Structures for Faster Nearest-Neighbor Classification(2022) Flores Velazco, Alejandro; Mount, David; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Given a set P of n labeled points in a metric space (X,d), the nearest-neighbor rule classifies an unlabeled query point q ∈ X with the class of q's closest point in P. Despite the advent of more sophisticated techniques, nearest-neighbor classification is still fundamental for many machine-learning applications. Over the years, this~has motivated numerous research aiming to reduce its high dependency on the size and dimensionality of the data. This dissertation presents various approaches to reduce the dependency of the nearest-neighbor rule from n to some smaller parameter k, that describes the intrinsic complexity of the class boundaries of P. This is of particular significance as it is usually assumed that k ≪ n on real-world training sets. One natural way to achieve this dependency reduction is to reduce the training set itself, selecting a subset R ⊆ P to be used by the nearest-neighbor rule~to~answer incoming queries, instead of using P. Evidently, this approach would reduce the dependencies of the nearest-neighbor rule from n, the size of P, to the size of R. This dissertation explores different techniques to select subsets whose sizes are proportional to k, and that provide varying degrees of correct classification guarantees. Another alternative involves bypassing training set reduction, and instead building data structures designed to answer classification queries directly. To this end, this dissertation proposes the Chromatic AVD; a Quadtree-based data structure designed to answer ε-approximate nearest-neighbor classification queries. The query time and space complexities of this data structure depend on k_ε; a generalization of k that describes the intrinsic complexity of the ε-approximate class boundaries of P.Item Rich and Scalable Models for Text(2019) nguyen, thang dai; Boyd-Graber, Jordan; Resnik, Philip; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Topic models have become essential tools for uncovering hidden structures in big data. However, the most popular topic model algorithm—Latent Dirichlet Allocation (LDA)— and its extensions suffer from sluggish performance on big datasets. Recently, the machine learning community has attacked this problem using spectral learning approaches such as the moment method with tensor decomposition or matrix factorization. The anchor word algorithm by Arora et al. [2013] has emerged as a more efficient approach to solve a large class of topic modeling problems. The anchor word algorithm is high-speed, and it has a provable theoretical guarantee: it will converge to a global solution given enough number of documents. In this thesis, we present a series of spectral models based on the anchor word algorithm to serve a broader class of datasets and to provide more abundant and more flexible modeling capacity. First, we improve the anchor word algorithm by incorporating various rich priors in the form of appropriate regularization terms. Our new regularized anchor word algorithms produce higher topic quality and provide flexibility to incorporate informed priors, creating the ability to discover topics more suited for external knowledge. Second, we enrich the anchor word algorithm with metadata-based word representation for labeled datasets. Our new supervised anchor word algorithm runs very fast and predicts better than supervised topic models such as Supervised LDA on three sentiment datasets. Also, sentiment anchor words, which play a vital role in generating sentiment topics, provide cues to understand sentiment datasets better than unsupervised topic models. Lastly, we examine ALTO, an active learning framework with a static topic overview, and investigate the usability of supervised topic models for active learning. We develop a new, dynamic, active learning framework that combines the concept of informativeness and representativeness of documents using dynamically updating topics from our fast supervised anchor word algorithm. Experiments using three multi-class datasets show that our new framework consistently improves classification accuracy over ALTO.Item FACE RECOGNITION AND VERIFICATION IN UNCONSTRAINED ENVIRIONMENTS(2012) Guo, Huimin; Davis, Larry; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Face recognition has been a long standing problem in computer vision. General face recognition is challenging because of large appearance variability due to factors including pose, ambient lighting, expression, size of the face, age, and distance from the camera, etc. There are very accurate techniques to perform face recognition in controlled environments, especially when large numbers of samples are available for each face (individual). However, face identification under uncontrolled( unconstrained) environments or with limited training data is still an unsolved problem. There are two face recognition tasks: face identification (who is who in a probe face set, given a gallery face set) and face verification (same or not, given two faces). In this work, we study both face identification and verification in unconstrained environments. Firstly, we propose a face verification framework that combines Partial Least Squares (PLS) and the One-Shot similarity model[1]. The idea is to describe a face with a large feature set combining shape, texture and color information. PLS regression is applied to perform multi-channel feature weighting on this large feature set. Finally the PLS regression is used to compute the similarity score of an image pair by One-Shot learning (using a fixed negative set). Secondly, we study face identification with image sets, where the gallery and probe are sets of face images of an individual. We model a face set by its covariance matrix (COV) which is a natural 2nd-order statistic of a sample set.By exploring an efficient metric for the SPD matrices, i.e., Log-Euclidean Distance (LED), we derive a kernel function that explicitly maps the covariance matrix from the Riemannian manifold to Euclidean space. Then, discriminative learning is performed on the COV manifold: the learning aims to maximize the between-class COV distance and minimize the within-class COV distance. Sparse representation and dictionary learning have been widely used in face recognition, especially when large numbers of samples are available for each face (individual). Sparse coding is promising since it provides a more stable and discriminative face representation. In the last part of our work, we explore sparse coding and dictionary learning for face verification application. More specifically, in one approach, we apply sparse representations to face verification in two ways via a fix reference set as dictionary. In the other approach, we propose a dictionary learning framework with explicit pairwise constraints, which unifies the discriminative dictionary learning for pair matching (face verification) and classification (face recognition) problems.Item Cost-sensitive Information Acquisition in Structured Domains(2010) Bilgic, Mustafa; Getoor, Lise C; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Many real-world prediction tasks require collecting information about the domain entities to achieve better predictive performance. Collecting the additional information is often a costly process that involves acquiring the features describing the entities and annotating the entities with target labels. For example, document collections need to be manually annotated for classification and lab tests need to be ordered for medical diagnosis. Annotating the whole document collection and ordering all possible lab tests might be infeasible due to limited resources. In this thesis, I explore effective and efficient ways of choosing the right features and labels to acquire under limited resources. For the problem of feature acquisition, we are given entities with missing features and the task is to classify them with minimum cost. The likelihood of misclassification can be reduced by acquiring features but acquiring features incurs costs as well. The objective is to acquire the right set of features that balance acquisition and misclassification cost. I introduce a technique that can reduce the space of possible sets of features to consider for acquisition by exploiting the conditional independence properties in the underlying probability distribution. For the problem of label acquisition, I consider two real-world scenarios. In the first one, we are given a previously trained model and a budget determining how many labels we can acquire, and the objective is to determine the right set of labels to acquire so that the accuracy on the remaining ones is maximized. I describe a system that can automatically learn and predict on which entities the underlying classifier is likely to make mistakes and it suggests acquiring the labels of the entities that lie in a high density potentially-misclassified region. In the second scenario, we are given a network of entities that are unlabeled and our objective is to learn a classification model that will have the least future expected error by acquiring minimum number of labels. I describe an active learning technique that can exploit the relationships in the network both to select informative entities to label and to learn a collective classifier that utilizes the label correlations in the network.