Research map / Learning theory and algorithms / Matrix and tensor methods
Elastic Functional Coding of Riemannian Trajectories
Rushil Anirudh, Pavan K. Turaga, Jingyong Su, Anuj Srivastava
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 |
|---|---|---|---|
| add TSRVF elastic warping instead of DTW (ablation) | LARP features with SVM · MSRActions3D (20 classes) | accuracy: 78.75 ± 3.08 -> 84.62 ± 3.08 | Improved |
| add TSRVF elastic warping instead of DTW (ablation) | LARP features with SVM · Florence3D (9 actions) | accuracy: 86.74 -> 89.50 | Improved |
| add TSRVF elastic warping instead of DTW (ablation) | LARP features with SVM · UTKinect (10 actions) | accuracy: 92.17 -> 94.47 | Improved |
| add RF-PCA coding on TSRVF shooting vectors (about 250x fewer dims) (ablation) | TSRVF · Florence3D (9 actions) | accuracy: 89.50 -> 89.67 | Improved |
| add RF-PCA coding on TSRVF shooting vectors (ablation) | TSRVF · MSRActions3D (20 classes) | accuracy: 84.62 ± 3.08 -> 85.16 ± 3.13 | Improved |
| use PCA on shooting vectors of whole trajectories instead of PGA on individual time samples (ablation) | LARP features · MSRActions3D (20 classes) | accuracy: 72.06 ± 3.12 -> 85.16 ± 3.13 | Improved |
| use RF-LCKSVD with a small dictionary (few samples per class) (ablation) | RF coding · MSRActions3D (20 classes) | accuracy: 85.16 ± 3.13 -> 83.60 ± 3.14 | Hurt |
| add RF-PCA coding to TSRVF shooting vectors with SVM (ablation) | TSRVF · OuluVS (SDT, 1000 videos) | accuracy: 74.05 ± 4.14 -> 75.3 ± 5.41 | Improved |
| use TSRVF + SVM instead of TSRVF + NN (ablation) | OuluVS covariance trajectories · OuluVS (SDT, 1000 videos) | accuracy: 66.0 -> 74.05 | Improved |
| add elasticity to SE(3) trajectories, then functional codes (ablation) | LARP features · Stroke rehabilitation WMFT | correlation with therapist score: 92.27 -> 93.53 -> 97.84 | Improved |
| choose an arbitrary reference point instead of the Riemannian mean (ablation) | TSRVF warping · UMD actions (10 similar actions) | registration convergence: about 35 iterations to converge vs no convergence after 80 iterations | Hurt |
| add RF-PCA coding under sensor noise (ablation) | TSRVF · UMD actions (Grassmann, noise) | accuracy | Improved |
Closest papers by meaning
- Learned Trajectory Embedding for Subspace Clustering (CVPR 2024 · similarity 0.86)
- Follow the Energy, Find the Path: Riemannian Metrics from Energy-Based Models (NeurIPS 2025 · similarity 0.86)
- Task-Induced Riemannian Metrics for Vision Transformer Feature Spaces (NeurIPS 2026 · similarity 0.85)
- Finsler Multi-Dimensional Scaling: Manifold Learning for Asymmetric Dimensionality Reduction and Embedding (CVPR 2025 · similarity 0.85)
- Data-free Neural Representation Compression with Riemannian Neural Dynamics (ICML 2024 · similarity 0.84)
- Bringing Motion Taxonomies to Continuous Domains via GPLVM on Hyperbolic manifolds (ICML 2024 · 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.
