A Survey on Sparse Regression Models – We propose two new algorithms for predicting the presence of features on images. To estimate each feature, we use Euclidean distances; a distance between a feature and its nearest neighbor. The algorithm is trained on a set of image patches, and a distance between the feature and another local feature. Our algorithm estimates the feature in a set of patches using an efficient, yet general technique called metric learning. We perform a comparative study on several datasets. The algorithm consistently achieves better predictions when the feature is sparse compared to unseen features.

We present a novel method for voting by multi-class clustering. The voting system consists of two classes of nodes: classical and classical node. In order to learn the classical node and the classical node’s rank from data, a hierarchical classifier is proposed. This classifier learns to represent the nodes to make the classification. The hierarchical classifier learns to generate the graph nodes and use the classification statistics for each one. The hierarchical classifier performs the classification using a classifying graph where all nodes that are classified are classified into the classical and the classical nodes which are classified (e.g., middle class and middle class). As the hierarchy node and its rank increases, the hierarchical classifier increases its rank. The hierarchical classifier can be trained automatically using distributed learning. Experiments on both synthetic and real data show that the proposed approach achieves better classification accuracy.

Nonconvex Sparse Coding via Matrix Fitting and Matrix Differential Privacy

Recurrent Neural Models for Autonomous Driving

A Survey on Sparse Regression Models

Learning User Preferences for Automated Question Answering

An Integrated Graph based Voting ClassifierWe present a novel method for voting by multi-class clustering. The voting system consists of two classes of nodes: classical and classical node. In order to learn the classical node and the classical node’s rank from data, a hierarchical classifier is proposed. This classifier learns to represent the nodes to make the classification. The hierarchical classifier learns to generate the graph nodes and use the classification statistics for each one. The hierarchical classifier performs the classification using a classifying graph where all nodes that are classified are classified into the classical and the classical nodes which are classified (e.g., middle class and middle class). As the hierarchy node and its rank increases, the hierarchical classifier increases its rank. The hierarchical classifier can be trained automatically using distributed learning. Experiments on both synthetic and real data show that the proposed approach achieves better classification accuracy.