The paper is concerned with two-class active learning. While the common approach for collecting data in active learning is to select samples close to the classification boundary, better performance can be achieved by taking into account the prior data distribution. The main contribution of the paper is a formal framework that incorporates clustering into active learning. The algorithm first constructs a classifier on the set of the cluster representatives, and then propagates the classification decision to the other samples via a local noise model. The proposed model allows to select the most representative samples as well as to avoid repeatedly labeling samples in the same cluster. During the active learning process, the clustering is adjusted using the coarse-to-fine strategy in order to balance between the advantage of large clusters and the accuracy of the data representation. The results of experiments in image databases show a better performance of our algorithm compared to the current methods.

Abstract Boosting is one of the most significant advances in machine learning for classification and regression. In its original and computationally flexible version, boosting seeks to minimize empirically a loss function in a greedy fashion. The resulted estimator takes an additive function form and is built iteratively by applying a base estimator (or learner) to updated samples depending on the previous iterations. An unusual regularization technique, early stopping, is employed based on CV or a test set. This paper studies numerical convergence, consistency, and statistical rates of convergence of boosting with early stopping, when it is carried out over the linear span of a family of basis functions. For general loss functions, we prove the convergence of boosting's greedy optimization to the infinimum of the loss function over the linear span. Using the numerical convergence result, we find early stopping strategies under which boosting is shown to be consistent based on iid samples, and we obtain bounds on the rates of convergence for boosting estimators. Simulation studies are also presented to illustrate the relevance of our theoretical results for providing insights to practical aspects of boosting. As a side product, these results also reveal the importance of restricting the greedy search step sizes, as known in practice through the works of Friedman and others. Moreover, our results lead to a rigorous proof that for a linearly separable problem, AdaBoost with ffl! 0 stepsize becomes an L1-margin maximizer when left to run to convergence. 1 Introduction In this paper we consider boosting algorithms for classification and regression. These algorithms present one of the major progresses in machine learning. In their original version, the computational aspect is explicitly specified as part of the estimator/algorithm. That is, the empirical minimization of an appropriate loss function is carried out in a greedy fashion, which means that at each step, a basis function that leads to the largest reduction of empirical risk is added into the estimator. This specification distinguishes boosting from other statistical procedures which are defined by an empirical minimization of a loss function without the numerical optimization details.

Real-world applications of text categorization often require a system to deal with tens of thousands of categories de- ned over a large taxonomy. This paper addresses the problem with respect to a set of popular algorithms in text categorization, including Support Vector Machines, k-nearest neighbor, ridge regression, linear least square t and logistic regression. By providing a formal analysis of the computational complexity of each classi cation method, followed by an investigation on the usage of dierent classi ers in a hierarchical setting of categorization, we show how the scalability of a method depends on the topology of the hierarchy and the category distributions. In addition, we are able to obtain tight bounds for the complexities by using the power law to approximate category distributions over a hierarchy. Experiments with kNN and SVM classi ers on the OHSUMED corpus are reported on, as concrete examples.

Abstract not found

To solve the problem of indexing collections with diverse text documents, image documents, or documents with both text and images, one needs to develop a model that supports heterogeneous types of documents. In this paper, we show how information theory supplies us with the tools necessary to develop a unique model for text, image, and text/image retrieval. In our approach, for each possible query keyword we estimate a maximum entropy model based on exclusively continuous features that were preprocessed. The unique continuous feature-space of text and visual data is constructed by using a minimum description length criterion to find the optimal feature-space representation (optimal from an information theory point of view). We evaluate our approach in three experiments: only text retrieval, only image retrieval, and text combined with image retrieval.

We study a setting that is motivated by the problem of filtering spam messages for many users. Each user receives messages according to an individual, unknown distribution, reflected only in the unlabeled inbox. The spam filter for a user is required to perform well with respect to this distribution. Labeled messages from publicly available sources can be utilized, but they are governed by a distinct distribution, not adequately representing most inboxes. We devise a method that minimizes a loss function with respect to a user’s personal distribution based on the available biased sample. A nonparametric hierarchical Bayesian model furthermore generalizes across users by learning a common prior which is imposed on new email accounts. Empirically, we observe that bias-corrected learning outperforms naive reliance on the assumption of independent and identically distributed data; Dirichlet-enhanced generalization across users outperforms a single (“one size fits all”) filter as well as independent filters for all users. 1

A number of feature selection metrics have been explored in text categorization, among which information gain (IG), chi-square (CHI), correlation coefficient (CC) and odds ratios (OR) are considered most effective. CC and OR are one-sided metrics while IG and CHI are two-sided. Feature selection using one-sided metrics selects the features most indicative of membership only, while feature selection using two-sided metrics implicitly combines the features most indicative of membership (e.g. positive features) and nonmembership (e.g. negative features) by ignoring the signs of features. The former never consider the negative features, which are quite valuable, while the latter cannot ensure the optimal combination of the two kinds of features especially on imbalanced data. In this work, we investigate the usefulness of explicit control of that combination within a proposed feature selection framework. Using multinomial naïve Bayes and regularized logistic regression as classifiers, our experiments show both great potential and actual merits of explicitly combining positive and negative features in a nearly optimal fashion according to the imbalanced data. 1.

This paper reports a cross-benchmark evaluation of regularized logistic regression (LR) and incremental Rocchio for adaptive filtering. Using four corpora from the Topic Detection and Tracking (TDT) forum and the Text Retrieval Conferences (TREC) we evaluated these methods with non-stationary topics at various granularity levels, and measured performance with different utility settings. We found that LR performs strongly and robustly in optimizing T11SU (a TREC utility function) while Rocchio is better for optimizing Ctrk (the TDT tracking cost), a high-recall oriented objective function. Using systematic cross-corpus parameter optimization with both methods, we obtained the best results ever reported on TDT5, TREC10 and TREC11. Relevance feedback on a small portion (0.05~0.2%) of the TDT5 test documents yielded significant performance improvements, measuring up to a 54 % reduction in Ctrk and a 20.9 % increase in T11SU (with β=0.1), compared to the results of the top-performing system in TDT2004 without relevance feedback information.

Classifiers favoring sparse solutions, such as support vector machines, relevance vector machines, LASSO-regression based classifiers, etc., provide competitive methods for classification problems in high dimensions. However, current algorithms for training sparse classifiers typically scale quite unfavorably with respect to the number of training examples. This paper proposes online and multi-pass algorithms for training sparse linear classifiers for high dimensional data. These algorithms have computational complexity and memory requirements that make learning on massive datasets feasible. The central idea that makes this possible is a straightforward quadratic approximation to the likelihood function.

The problem that this thesis concentrates on is phishing attacks. Phishing attacks use email messages and web sites designed to look as if they come from a known and legitimate organization, in order to deceive users into submitting their personal, financial, or computer account information online at those fake web sites. Phishing is a semantic attack. The fundamental problem of phishing is that when a user submits sensitive information online under an attack, his mental model about this submission is different from the system model that actually performs this submission. Specifically, the system sends the data to a different web site from the one where the user intends to submit the data. The fundamental solution to phishing is to bridge the semantic gap between the user’s mental model and the system model. The user interface is where human users interact with the computer system. It is where a user’s intention transforms into a system operation. It is where the semantic gap happens under phishing attacks. And therefore, it is where the phishing should be solved.