Optimizing Rank-based Metrics with Blackbox Differentiation

Abstract

Rank-based metrics are some of the most widely used criteria for performance evaluation of computer vision models. Despite years of effort, direct optimization for these metrics remains a challenge due to their non-differentiable and non-decomposable nature. We present an efficient, theoretically sound, and general method for differentiating rank-based metrics with mini-batch gradient descent. In addition, we address optimization instability and sparsity of the supervision signal that both arise from using rank-based metrics as optimization targets. Resulting losses based on recall and Average Precision are applied to image retrieval and object detection tasks. We obtain performance that is competitive with state-of-the-art on standard image retrieval datasets and consistently improve performance of near state-of-the-art object detectors. The code is available at https://github.com/martius-lab/blackbox-backprop

arXiv commentCVPR 2020 conference paper (oral). The first two authors contributed equally

References in corpus (9)

• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• FaceNet: A Unified Embedding for Face Recognition and Clustering
• YOLOv3: An Incremental Improvement
• MMDetection: Open MMLab Detection Toolbox and Benchmark
• A Survey on Metric Learning for Feature Vectors and Structured Data
• Learning Deep Embeddings with Histogram Loss
• Towards Accurate One-Stage Object Detection with AP-Loss

Cited by in corpus (13)

• Decision-Focused Learning: Foundations, State of the Art, Benchmark and Future Opportunities
• Deep Graph Matching via Blackbox Differentiation of Combinatorial Solvers
• HOPE: Hybrid-granularity Ordinal Prototype Learning for Progression Prediction of Mild Cognitive Impairment
• Fast Differentiable Sorting and Ranking
• LibAUC: A Deep Learning Library for X-Risk Optimization
• Stochastic Optimization of Areas Under Precision-Recall Curves with Provable Convergence
• Robust and Decomposable Average Precision for Image Retrieval
• Set-to-Sequence Methods in Machine Learning: a Review
• Decision-Oriented Learning for Future Power System Decision-Making under Uncertainty
• Learning with Algorithmic Supervision via Continuous Relaxations
• Momentum Accelerates the Convergence of Stochastic AUPRC Maximization
• Deep AUC Maximization for Medical Image Classification: Challenges and Opportunities
• Combinatorial Optimization for Panoptic Segmentation: A Fully Differentiable Approach