CVAug 25, 2021

Recall@k Surrogate Loss with Large Batches and Similarity Mixup

arXiv:2108.11179v270 citations
Originality Highly original
AI Analysis

This work addresses the challenge of directly optimizing recall for retrieval tasks, which is incremental as it builds on existing metric learning approaches.

The paper tackles the problem of optimizing non-differentiable recall metrics in deep visual retrieval by proposing a differentiable surrogate loss, enabling training with large batch sizes and similarity mixup regularization. It achieves state-of-the-art performance in image retrieval benchmarks, outperforming methods using average precision approximations.

This work focuses on learning deep visual representation models for retrieval by exploring the interplay between a new loss function, the batch size, and a new regularization approach. Direct optimization, by gradient descent, of an evaluation metric, is not possible when it is non-differentiable, which is the case for recall in retrieval. A differentiable surrogate loss for the recall is proposed in this work. Using an implementation that sidesteps the hardware constraints of the GPU memory, the method trains with a very large batch size, which is essential for metrics computed on the entire retrieval database. It is assisted by an efficient mixup regularization approach that operates on pairwise scalar similarities and virtually increases the batch size further. The suggested method achieves state-of-the-art performance in several image retrieval benchmarks when used for deep metric learning. For instance-level recognition, the method outperforms similar approaches that train using an approximation of average precision.

Code Implementations2 repos
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes