Research map / Label-efficient and robust learning / Recognition and long-tail
OLÉ: Orthogonal Low-Rank Embedding - A Plug and Play Geometric Loss for Deep 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 |
|---|---|---|---|
| Add OLÉ loss to softmax training | DenseNet-40-12 · SVHN test | error (%): 3.93± 0.08 -> 3.62± 0.04 | Helped |
| Add OLÉ loss to softmax training | DenseNet-40-12 · MNIST test | error (%): 0.88± 0.03 -> 0.78± 0.04 | Helped |
| Add OLÉ loss to softmax training | ResNet-110 · CIFAR10+ test | error (%): 6.05± 0.8 -> 5.39± 0.25 | Helped |
| Add OLÉ loss to softmax training | VGG-16 · CIFAR10 test | error (%): 8.23± 0.13 -> 7.22± 0.14 | Helped |
| Add OLÉ loss to softmax training | PreResNet-110 · CIFAR100+ test | error (%): 23.01± 0.19 -> 22.8± 0.34 | Helped |
| Add OLÉ loss to softmax training | VGG-FACE · Facescrub-500 test | error (%): 2.49± 0.01 -> 1.55± 0.02 | Helped |
| Add OLÉ loss to softmax training | CNN-5 · STL-10+ test | error (%): 18.22 ± 0.27 -> 16.68± 0.24 | Helped |
| Add OLÉ loss to softmax training | VGG-FACE · LFW | AUC (%): 99.12% vs $99.04\%$ | No clear effect |
| Use OLÉ standalone instead of softmax loss | 3-layer MLP on Facescrub identities · Facescrub illustrative classification | classification performance | Helped |
| Use OLÉ standalone instead of softmax loss | 4-layer MLP on CIFAR10 classes · CIFAR10 illustrative classification | classification performance | Helped |
| Add OLÉ loss to softmax training | CNN-5 · STL-10, varying training samples per class | error (%) | Helped |
Closest papers by meaning
- Improved Deep Metric Learning with Multi-class N-pair Loss Objective (NeurIPS 2016 · similarity 0.88)
- Learning Towards The Largest Margins (ICLR 2022 · similarity 0.88)
- SSOLE: Rethinking Orthogonal Low-rank Embedding for Self-Supervised Learning (ICLR 2025 · similarity 0.87)
- ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019 · similarity 0.86)
- Dual Pose-invariant Embeddings: Learning Category and Object-specific Discriminative Representations for Recognition and Retrieval (CVPR 2024 · similarity 0.86)
- Unlocking Better Closed-Set Alignment Based on Neural Collapse for Open-Set Recognition (AAAI 2025 · similarity 0.85)
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.