Research map / Learning theory and algorithms / Clustering
Graph Constrained Data Representation Learning for Human Motion Segmentation
Mariella Dimiccoli, Lluís Garrido, Guillem Rodriguez-Corominas, Herwig Wendt
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 auxiliary data matrix with graph constrained loss (ablation) | temporal subspace clustering (TSC-style) model · Weiz | NMI: 0.8277 -> 0.9053 | Improved |
| add auxiliary data matrix with graph constrained loss (ablation) | temporal subspace clustering (TSC-style) model · Keck | NMI: 0.8000 -> 0.8325 | Improved |
| add auxiliary data matrix with graph constrained loss (ablation) | temporal subspace clustering (TSC-style) model · UT | NMI: 0.8176 -> 0.8256 | Improved |
| add auxiliary data matrix with graph constrained loss (ablation) | temporal subspace clustering (TSC-style) model · MAD | NMI: 0.7529 -> 0.8471 | Improved |
| decrease relative weight of graph loss L_G by factor 10 (ablation) | Ours · Weiz subject 1 | smoothness of auxiliary data: smoothness increased by a factor 10^0.7 ≈ 5-8 at 10^0.05 ≈ 10-15% increased deviation | Improved |
| increase lambda_2 by two orders of magnitude (ablation) | Ours · Weiz subject 1 | smoothness of auxiliary data: smoothness 10^0.3 ≈ 2 at more than 10^0.08 ≈ 20% larger deviation | Mixed |
| use graph constrained auxiliary data under added Gaussian noise vs TSC and LTS on raw data | Ours · Weiz subject 1 (noisy frames) | NMI | Improved |
| set temporal regularization lambda_2 very small (ablation) | Ours · Weiz | NMI | Hurt |
| vary block-diagonal weight lambda_1 (ablation) | Ours · Weiz | NMI | No clear effect |
| vary graph filtering parameter h over two orders of magnitude (ablation) | Ours · Weiz | NMI | No clear effect |
Closest papers by meaning
- Learning Transferable Subspace for Human Motion Segmentation (AAAI 2018 · similarity 0.95)
- Probabilistic Temporal Subspace Clustering (CVPR 2017 · similarity 0.91)
- Consistency and Diversity Induced Human Motion Segmentation (TPAMI 2022 · similarity 0.91)
- Learned Trajectory Embedding for Subspace Clustering (CVPR 2024 · similarity 0.88)
- Learning Segmentation from Point Trajectories (NeurIPS 2024 · similarity 0.87)
- Exploring a Principled Framework for Deep Subspace Clustering (ICLR 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.
