Research map / Optimization / Optimizers
Siamese Networks: The Tale of Two Manifolds
Soumava Kumar Roy, Mehrtash Harandi, Richard Nock, Richard Hartley
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 qConv layer instead of ordinary Conv embedding layer (ablation) | Siamese network · MNIST | 1-NN error (%): 14.84 -> 9.93 | Improved |
| use Stiefel layer instead of ordinary Conv embedding layer (ablation) | Siamese network · MNIST | 1-NN error (%): 14.84 -> 9.75 | Improved |
| use qConv layer instead of ordinary Conv embedding layer (ablation) | shallow Siamese network · STL10 | NN accuracy (%): 42.7% -> 48.5% | Improved |
| use Stiefel layer instead of ordinary Conv embedding layer (ablation) | shallow Siamese network · STL10 | NN accuracy (%): 42.7% -> 48.6% | Improved |
| add Riemannian geometry (Stiefel) to triplet-trained Inception (ablation) | Trip-SH · CARS196 | Recall@1: 51.5 -> 73.2 | Improved |
| add Riemannian geometry (Stiefel) to triplet-trained Inception (ablation) | Trip-SH · CUB-200-2011 | Recall@1: 42.6 -> 52.3 | Improved |
| add Riemannian geometry (Stiefel) to triplet-trained Inception (ablation) | Trip-SH · SOP | Recall@1: 66.7 -> 69.2 | Improved |
| Stiefel geometry versus recent complex methods DAML/DVML/DWS on CUB | Ours (Stiefel) · CUB-200-2011 | Recall@1: 52.3 vs 52.7 (DVML) | No clear effect |
| qConv training time overhead versus Conv per epoch (ablation) | qConv · MNIST | time per epoch (sec): 12.1 -> 12.3 | No clear effect |
| qConv and Stiefel training time overhead versus Conv (ablation) | Siamese network · STL10 | time per epoch (sec): 143 -> 150 and 152 | No clear effect |
Closest papers by meaning
- ManifoldNet: A Deep Neural Network for Manifold-Valued Data With Applications (TPAMI 2020 · similarity 0.88)
- Riemannian Local Mechanism for SPD Neural Networks (AAAI 2023 · similarity 0.88)
- Exploring Simple Siamese Representation Learning (CVPR 2021 · similarity 0.88)
- A Statistical Recurrent Model on the Manifold of Symmetric Positive Definite Matrices (NeurIPS 2018 · similarity 0.88)
- Consistent Geometric Deep Learning via Hilbert Bundles and Cellular Sheaves (NeurIPS 2026 · similarity 0.86)
- Building Transformation Layers for Riemannian Neural Networks (NeurIPS 2026 · 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.
