Research map / Optimization / Optimizers
Denoising Multi-Similarity Formulation: A Self-Paced Curriculum-Driven Approach for Robust Metric Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use BSPML instead of the strongest listed baseline | Metric learning · CUB | Recall@1: 49.6 -> 50.6 | Helped |
| Use BSPML instead of the strongest listed baseline | Metric learning · Cars-196 | Recall@1: 58.5 -> 59.6 | Helped |
| Use BSPML instead of the strongest listed baseline | Metric learning · In-Shop | Recall@1: 74.2 -> 75.0 | Helped |
| Use BSPML instead of the strongest listed baseline | Metric learning · SOP | Recall@1: 66.3 -> 66.3 | No clear effect |
| Use self-paced weighting to exclude noisy samples | Multi-Similarity metric learning · CUB, 10–30% noise | Recall@1: 48.46±0.16 -> 50.08±0.18; 47.79±0.19 -> 49.38±0.20; 46.66±0.15 -> 48.74±0.19 | Helped |
| Use self-paced weighting to exclude noisy samples | Multi-Similarity metric learning · Cars-196, 10–30% noise | Recall@1: 57.84±0.15 -> 58.78±0.18; 57.18±0.15 -> 58.24±0.16; 56.06±0.16 -> 57.38±0.19 | Helped |
| Use self-paced weighting to exclude noisy samples | Multi-Similarity metric learning · In-Shop, 10–30% noise | Recall@1: 70.51±0.17 -> 71.44±0.23; 66.38±0.17 -> 67.91±0.25; 60.25±0.15 -> 62.04±0.23 | Helped |
| Add self-paced learning to MS loss, without balance regularization (ablation) | MS loss · Noisy data sets; embedding sizes 64, 128, 256, 512 | Recall@1 | Helped |
| Add the balance regularization term to SPMS (ablation) | SPMS · Noisy data sets; embedding sizes 64, 128, 256, 512 | Recall@1 | Helped |
| Increase the age parameter λ (ablation) | BSPML · Noisy data sets | Mean of Average Weights of classes | Helped |
| Increase balance regularization μ (ablation) | BSPML · Noisy data sets with 20% noisy samples | Standard Deviation of Average Weights of classes | Helped |
Closest papers by meaning
- Noise-Resistant Deep Metric Learning With Ranking-Based Instance Selection (CVPR 2021 · similarity 0.88)
- Signal-To-Noise Ratio: A Robust Distance Metric for Deep Metric Learning (CVPR 2019 · similarity 0.88)
- Revisiting Training Strategies and Generalization Performance in Deep Metric Learning (ICML 2020 · similarity 0.88)
- Large-Scale Distance Metric Learning With Uncertainty (CVPR 2018 · similarity 0.87)
- Deep Compositional Metric Learning (CVPR 2021 · similarity 0.87)
- DarkRank: Accelerating Deep Metric Learning via Cross Sample Similarities Transfer (AAAI 2018 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.