1,721,042 research outputs found
Discriminative Learning via Semidefinite Probabilistic Models
No full textDiscriminative linear models are a popular tool in machine learning. These can be generally divided into two types: linear classifiers, such as support vector machines (SVMs), which are well studied and provide stateof-the-art results, and probabilistic models such as logistic regression. One shortcoming of SVMs is that their output (known as the ”margin”) is not calibrated, so that it is difficult to incorporate such models as components of larger systems. This problem is solved in the probabilistic approach. We combine these two approaches above by constructing a model which is both linear in the model parameters and probabilistic, thus allowing maximum margin training with calibrated outputs. Our model assumes that classes correspond to linear subspaces (rather than to half spaces), a view which is closely related to concepts in quantum detection theory. 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 second-order kernel methods.
Efficient online learning with individual learning-rates for phoneme sequence recognition
No full textGeneral Terms
No full textWe describe a new family of topic-ranking algorithms for multi-labeled documents. The motivation for the algorithms stems from recent advances in online learning algorithms. The algorithms we present are simple to implement and are time and memory efficient. We evaluate the algorithms on the Reuters-21578 corpus and the new corpus released by Reuters in 2000. On both corpora the algorithms we present outperform adaptations to topic-ranking of Rocchio’s algorithm and the Perceptron algorithm. We also outline the formal analysis of the algorithm in the mistake bound model. To our knowledge, this work is the first to report performance results with the entire new Reuters corpus
Learning from Data of Variable Quality
No full textWe initiate the study of learning from multiple sources of limited data, each of which may be corrupted at a different rate. We develop a complete theory of which data sources should be used for two fundamental problems: estimating the bias of a coin, and learning a classifier in the presence of label noise. In both cases, efficient algorithms are provided for computing the optimal subset of data
- …
