Non-Convex Robust Low-Rank Matrix Estimation via Subspace Learning – We present a novel strategy to train low-rank matrix estimators from low-rank matrices. The model is based on convex optimization of the projection matrix, which is a common optimization task in many image understanding applications. Since many matrices that support the projection matrix cannot be represented in a constrained space, this strategy requires the sparse matrix to be added. We show that sparse projection matrices can be efficiently learned via the convex optimization algorithm. This allows our algorithm to learn a low-rank matrix with a non-convex loss. We show that it also can be learned with a sparse matrices. We provide a simple algorithmic analysis for our algorithm for solving the convex optimization problem.
We propose a novel and practical method to classify road signs. The dataset comprises a 3D vehicle mounted vehicle system (VVST) and two navigation tasks, which are: (1) classification of road signs and (2) classification of vehicles. The vehicles are grouped into two classes, the sign classifier and the vehicle classifier. To classify road signs, we first learn a distance matrix of distances between two classes and then the rank of the road signs is estimated using a distance metric. Then an algorithm is applied to classify the sign classifier by training the sign classifier on a dataset of real road vehicles. In this paper, we will discuss the results.
Efficient Geodesic Regularization on Graphs and Applications to Deep Learning Neural Networks
Dense Learning for Robust Road Traffic Speed Prediction
Non-Convex Robust Low-Rank Matrix Estimation via Subspace Learning
Dynamic Metric Learning with Spatial Neural Networks
A Multi-View Hierarchical Clustering Framework for Optimal Vehicle RoutingWe propose a novel and practical method to classify road signs. The dataset comprises a 3D vehicle mounted vehicle system (VVST) and two navigation tasks, which are: (1) classification of road signs and (2) classification of vehicles. The vehicles are grouped into two classes, the sign classifier and the vehicle classifier. To classify road signs, we first learn a distance matrix of distances between two classes and then the rank of the road signs is estimated using a distance metric. Then an algorithm is applied to classify the sign classifier by training the sign classifier on a dataset of real road vehicles. In this paper, we will discuss the results.