Distance Metric Learning Using Dropout: A Structured Regularization Approach

Distance metric learning (DML) aims to learn a distance metric better than Euclidean distance. It has been successfully applied to various tasks, e.g., classification, clustering and information retri