Optimizing Rank-based Metrics
RaMBO -- Rank Metric Blackbox Optimization
In search for practical application of the blackbox-differentiation theory, we turn to computer vision. Concretely, we show that applying blackbox-backprop to computer vision benchmarks in recall and Average Precision for retrieval and detection tasks consistently improves the underlying architectures’ performance.
The main component that enables this is the blackbox formulation of the argsort operation used for ranking, making the use of blackbox-differentiation theory possible. We made a blog post describing the method, which we call RaMBO (Rank Metric Blackbox Optimization). Further information about the paper (including a short and long oral presented at CVPR 2020) can be found here.
Publications
Optimizing Rank-based Metrics with Blackbox Differentiation
Conference on Computer Vision and Pattern Recognition
·
2020