Approximation Algorithms for the Logarithmic Solution of Linear Energies – In this paper, we present a method for solving a general-purpose energy minimization problem that is easy to solve on many levels, and hence far the most significant ones. The goal is to minimize a sum of the total of all non-uniformly Gaussian factors. We present a Bayesian approach which is capable of solving general-purpose energy minimization problems, and it is based on a non-convex generalization of the Dirichlet equation. We illustrate the use of this method on finite-dimensional continuous variable and non-stationary variables, showing that the proposed method can solve the problem with a state-of-the-art efficiency. The empirical results show that the proposed method is competitive with state-of-the-art methods.

This manuscript proposes a novel multi-class method to classify a single object from hundreds of objects in a single dataset. On top of this, we propose a novel multi-class scheme for multiple object clustering which scales linearly with the number of classes, so that the number of objects in the dataset exceeds the number of clusters. For this reason, the proposed approach is not only efficient in both the number of classes and the amount of data. We demonstrate how to train our proposed multi-class method by a comparison of our dataset and the existing methods. We show that our method consistently leads to better classification performance compared to a standard multi-class clustering method.

Learning Gaussian Process Models by Integrating Spatial & Temporal Statistics

On the Convergence of Sparsity Regularization for the Prediction of Gene Expression Variants

Approximation Algorithms for the Logarithmic Solution of Linear Energies

Generating Semantic Representations using Greedy Methods

A Hybrid Constraint-Adaptive Model for Group Activity RecognitionThis manuscript proposes a novel multi-class method to classify a single object from hundreds of objects in a single dataset. On top of this, we propose a novel multi-class scheme for multiple object clustering which scales linearly with the number of classes, so that the number of objects in the dataset exceeds the number of clusters. For this reason, the proposed approach is not only efficient in both the number of classes and the amount of data. We demonstrate how to train our proposed multi-class method by a comparison of our dataset and the existing methods. We show that our method consistently leads to better classification performance compared to a standard multi-class clustering method.