• Publications
  • Influence
Maximum Margin Clustering
TLDR
We propose a new method for clustering based on finding maximum margin hyperplanes through data. Expand
  • 464
  • 87
  • PDF
Probabilities for SV Machines
This chapter contains sections titled: Introduction, Fitting a Sigmoid After the SVM, Empirical Tests, Conclusions, Appendix: Pseudo-code for the Sigmoid Training
  • 422
  • 33
Reward Augmented Maximum Likelihood for Neural Structured Prediction
TLDR
We show that an optimal regularized expected reward is achieved when the conditional distribution of the outputs given the inputs is proportional to their exponentiated scaled rewards. Expand
  • 146
  • 29
  • PDF
Augmenting Naive Bayes Classifiers with Statistical Language Models
TLDR
We augment naive Bayes models with statistical n-gram language models to address short-comings of the standard naïve Bayes text classifier. Expand
  • 262
  • 26
  • PDF
Boosting in the Limit: Maximizing the Margin of Learned Ensembles
TLDR
We show that no simple version of the minimum-margin story can be complete, and show that Adaboost sometimes does overfit--eventually. Expand
  • 266
  • 25
  • PDF
Discriminative Batch Mode Active Learning
TLDR
We propose a new discriminative batch mode active learning strategy that exploits information from an unlabeled set to attempt to learn a good classifier directly. Expand
  • 253
  • 24
  • PDF
Bridging the Gap Between Value and Policy Based Reinforcement Learning
TLDR
We establish a new connection between value and policy based reinforcement learning (RL) based on a relationship between softmax temporal value consistency and policy optimality under entropy regularization. Expand
  • 208
  • 21
  • PDF
General Convergence Results for Linear Discriminant Updates
TLDR
We give a single proof of convergence that covers a broad subset of algorithms in this class, including both Perceptron and Winnow, but also many new algorithms. Expand
  • 164
  • 21
  • PDF
Constraint-based optimization and utility elicitation using the minimax decision criterion
TLDR
We address two problems associated with preference elicitation: computing a best feasible solution when the user's utilities are imprecisely specified; and developing useful elicitation procedures that reduce utility uncertainty, with minimal user interaction, to a point where optimal decisions can be made. Expand
  • 137
  • 20
  • PDF
Semi-Supervised Conditional Random Fields for Improved Sequence Segmentation and Labeling
TLDR
We present a new semi-supervised training procedure for conditional random fields (CRFs) that can be used to train sequence segmentors and labelers from a combination of labeled and unlabeled training data. Expand
  • 150
  • 19
  • PDF