arXiv subjects

Koby Crammer

Publications and source records attributed to Koby Crammer.

22 records · Page 2Linked to original sources

arXiv2012

More Is Better: Large Scale Partially-supervised Sentiment Classification - Appendix

We describe a bootstrapping algorithm to learn from partially labeled data, and the results of an empirical study for using it to improve performance of sentiment classification using up to 15 million unlabeled Amazon product reviews. Our experiments cover semi-supervised learning, domain adaptation and weakly supervised learning. In some cases our methods were able to reduce test error by more than half using such large amount of data. NOTICE: This is only the supplementary material.

cs.LG↗

arXiv2012

Discriminative Learning via Semidefinite Probabilistic Models

Discriminative linear models are a popular tool in machine learning. These can be generally divided into two types: The first is linear classifiers, such as support vector machines, which are well studied and provide state-of-the-art results. One shortcoming of these models is that their output (known as the 'margin') is not calibrated, and cannot be translated naturally into a distribution over the labels. Thus, it is difficult to incorporate such models as components of larger systems, unlike probabilistic based approaches. The second type of approach constructs class conditional distributions using a nonlinearity (e.g. log-linear models), but is occasionally worse in terms of classification error. We propose a supervised learning method which combines the best of both approaches. Specifically, our method provides a distribution over the labels, which is a linear function of the model parameters. As a consequence, differences between probabilities are linear functions, a property which most probabilistic models (e.g. log-linear) do not have. Our model assumes that classes correspond to linear subspaces (rather than to half spaces). Using a relaxed projection operator, we construct a measure which evaluates the degree to which a given vector 'belongs' to a subspace, resulting in a distribution over labels. Interestingly, this view is closely related to similar concepts in quantum detection theory. The resulting models can be trained either to maximize the margin or to optimize average likelihood measures. The corresponding optimization problems are semidefinite programs which can be solved efficiently. We illustrate the performance of our algorithm on real world datasets, and show that it outperforms 2nd order kernel methods.

cs.LG↗

arXiv2012

Adaptive Regularization for Weight Matrices

Algorithms for learning distributions over weight-vectors, such as AROW were recently shown empirically to achieve state-of-the-art performance at various problems, with strong theoretical guaranties. Extending these algorithms to matrix models pose challenges since the number of free parameters in the covariance of the distribution scales as $n^4$ with the dimension $n$ of the matrix, and $n$ tends to be large in real applications. We describe, analyze and experiment with two new algorithms for learning distribution of matrix models. Our first algorithm maintains a diagonal covariance over the parameters and can handle large covariance matrices. The second algorithm factors the covariance to capture inter-features correlation while keeping the number of parameters linear in the size of the original matrix. We analyze both algorithms in the mistake bound model and show a superior precision performance of our approach over other algorithms in two tasks: retrieving similar images, and ranking similar documents. The factored algorithm is shown to attain faster convergence rate.

cs.LG↗

arXiv2011

Confidence Estimation in Structured Prediction

Structured classification tasks such as sequence labeling and dependency parsing have seen much interest by the Natural Language Processing and the machine learning communities. Several online learning algorithms were adapted for structured tasks such as Perceptron, Passive- Aggressive and the recently introduced Confidence-Weighted learning . These online algorithms are easy to implement, fast to train and yield state-of-the-art performance. However, unlike probabilistic models like Hidden Markov Model and Conditional random fields, these methods generate models that output merely a prediction with no additional information regarding confidence in the correctness of the output. In this work we fill the gap proposing few alternatives to compute the confidence in the output of non-probabilistic algorithms.We show how to compute confidence estimates in the prediction such that the confidence reflects the probability that the word is labeled correctly. We then show how to use our methods to detect mislabeled words, trade recall for precision and active learning. We evaluate our methods on four noun-phrase chunking and named entity recognition sequence labeling tasks, and on dependency parsing for 14 languages.

cs.LG↗