To classify time series by nearest neighbor, we need to specify or learn a distance. We consider several variations of the Mahalanobis distance and the related Large Margin Nearest Neighbor Classification (LMNN). We find that the conventional Mahalanobis distance is counterproductive. However, both LMNN and the class-based diagonal Mahalanobis distance are competitive.

Current statistical models for structured prediction make simplifying assumptions about the underlying output graph structure, such as assuming a low-order Markov chain, because exact inference becomes intractable as the tree-width of the underlying graph increases. Approximate inference algorithms, on the other hand, force one to trade off representational power with computational efficiency. ...

The nearest neighbor classification is a simple and yet effective technique for pattern recognition. Performance of this technique depends significantly on the distance function used to compute similarity between examples. Some techniques were developed to learn weights of features for changing the distance structure of samples in nearest neighbor classification. In this paper, we propose an ap...

We have recently proposed an extension of ADABOOST to regression that uses the median of the base regressors as the final regressor. In this paper we extend theoretical results obtained for ADABOOST to median boosting and to its localized variant. First, we extend recent results on efficient margin maximizing to show that the algorithm can converge to the maximum achievable margin within a pres...

Guanidine hydrochloride has been widely used in the initial recovery steps of active protein from the inclusion bodies in aqueous two-phase system (ATPS). The knowledge of the guanidine hydrochloride effects on the liquid-liquid equilibrium (LLE) phase diagram behavior is still inadequate and no comprehensive theory exists for the prediction of the experimental trends. Therefore the effect the ...

Metric learning has been shown to significantly improve the accuracy of k-nearest neighbor (kNN) classification. In problems involving thousands of features, distance learning algorithms cannot be used due to overfitting and high computational complexity. In such cases, previous work has relied on a two-step solution: first apply dimensionality reduction methods to the data, and then learn a me...

We present a method to simultaneously learn a mixture of mappings and large margin hyperplane classifier. This method learns useful mappings of the training data to improve classification accuracy. We first present a simple iterative algorithm that finds a greedy local solution and then derive a semidefinite relaxation to find an approximate global solution. This relaxation leads to the matrix ...

We present a novel approach to collaborative prediction, using low-norm instead of low-rank factorizations. The approach is inspired by, and has strong connections to, large-margin linear discrimination. We show how to learn low-norm factorizations by solving a semi-definite program, and present generalization error bounds based on analyzing the Rademacher complexity of low-norm factorizations.