Multi proxy anchor family loss for several types of gradients

The deep metric learning (DML) objective is to learn a neural network that maps into an embedding space where similar data are near and dissimilar far. However, conventional proxy-based losses for DML have two problems: gradient problem application of the real-world dataset with multiple local centers. Additionally, performance metrics also some issues stability flexibility. This paper proposes three multi-proxies anchor (MPA) family normalized discounted cumulative gain (nDCG@k) metric. makes contributions. (1) MPA-family can using multi-local (2) improve training capacity owing solving problem. (3) data-wise or class-wise characteristics respect generation. Finally, we demonstrate effectiveness losses, achieves higher accuracy on datasets fine-grained images.