Research map / Architectures and efficiency / RNNs and normalization
Kissing to Find a Match: Efficient Low-Rank Permutation Representation
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 |
|---|---|---|---|
| Replace the shape-matching pipeline's permutation calculation with the low-rank factorization and Softmax | Marin et al. shape-matching pipeline · FAUST noisy | error1 (geodesic distance): 0.051 -> 0.047 | Helped |
| Replace the shape-matching pipeline's permutation calculation with the low-rank factorization and Softmax | Marin et al. shape-matching pipeline · FAUST noisy | error2 (geodesic distance): 0.029 -> 0.026 | Helped |
| Replace the shape-matching pipeline's permutation calculation with the low-rank factorization and Softmax | Marin et al. shape-matching pipeline · FAUST noise-free | error1 (geodesic distance): 0.043 -> 0.041 | Helped |
| Replace the shape-matching pipeline's permutation calculation with the low-rank factorization and Softmax | Marin et al. shape-matching pipeline · FAUST noise-free | error2 (geodesic distance): 0.022 -> 0.019 | Helped |
| Use stochastic training instead of computing the full permutation matrix (ablation) | Shape-matching network · FAUST evaluation after SURREAL training | error2 (geodesic distance): relative increase of less than 17% | Hurt |
| Use stochastic training instead of computing the full permutation matrix (ablation) | Shape-matching network · FAUST evaluation after SURREAL training | error1 (geodesic distance) | No clear effect |
| Use stochastic training with k=1 rather than full matrices (ablation) | Shape-matching network · TOSCA shape matching, 20000 vertices | memory usage: \sim73.6~\text{GB} -> 10.7~\text{GB} | Helped |
| Replace ReLU with Softmax in the low-rank permutation representation (ablation) | Point cloud alignment · n in {10,100,1000,10000} | correct nearest-neighbor permutation | No clear effect |
Closest papers by meaning
- OT4P: Unlocking Effective Orthogonal Group Path for Permutation Relaxation (NeurIPS 2024 · similarity 0.86)
- Learnable Permutation for Structured Sparsity on Transformer Models (AAAI 2026 · similarity 0.86)
- From Sorting Algorithms to Scalable Kernels: Bayesian Optimization in High-Dimensional Permutation Spaces (ICLR 2026 · similarity 0.86)
- Fast, Accurate and Memory-Efficient Partial Permutation Synchronization (CVPR 2022 · similarity 0.85)
- Polynomial Time Learning Augmented Algorithms for NP-hard Permutation Problems (ICML 2025 · similarity 0.84)
- Differentiable extensions with rounding guarantees for combinatorial optimization over permutations (NeurIPS 2025 · similarity 0.84)
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.