Research map / Time series / Causal and generative time series
Shape Distributions of Nonlinear Dynamical Systems for Video-Based Inference
Vinay Venkataraman, Pavan K. Turaga
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 |
|---|---|---|---|
| replace chaotic invariants with DT2 shape distribution (chi2) | phase space features · Motion capture | accuracy (%): 83.54 -> 99.37 | Improved |
| replace chaotic invariants with DT2 shape feature (linear SVM) | phase space features · MSR Action3D cross-subject | avg accuracy (%): 66.83 -> 87.89 | Improved |
| replace chaotic invariants with DT2 shape feature (NN classifier) | phase space features · MSR Action3D cross-subject | avg accuracy (%): 49.43 -> 79.25 | Improved |
| use chi-squared distance instead of Euclidean for DT2 (ablation) | DT2 · Motion capture | accuracy (%): 96.84 -> 99.37 | Improved |
| use chi-squared instead of Euclidean distance for D2 (ablation) | D2 · Motion capture | accuracy (%): 96.84 -> 96.84 | No clear effect |
| use local (time-windowed) shape function DT2 instead of global D2 (ablation) | shape distribution · MSR Action3D cross-subject (linear SVM) | avg accuracy (%): 86.12 -> 87.89 | Improved |
| increase embedding dimension from m=3 to m=5 for DT2 (ablation) | DT2 · MSR Action3D cross-subject (linear SVM) | avg accuracy (%): 87.89 -> 85.34 | Hurt |
| increase embedding dimension from m=3 to m=5 for DT2 (chi2) (ablation) | DT2 · Motion capture | accuracy (%): 99.37 -> 99.37 | No clear effect |
| replace KIM kinematic features with DT2 | stroke classification · Stroke rehabilitation | classification rate (%): 85.2 -> 92.05 | Improved |
| use D1 instead of kinematics KIM | stroke classification · Stroke rehabilitation | classification rate (%): 85.2 -> 84.32 | Hurt |
| use single wrist marker shape feature instead of kinematics on same marker | movement quality score · WMFT correlation | Pearson correlation: 0.6481 -> 0.8527 | Improved |
| replace KIM with DT2 for therapist-rated quality | movement quality · Stroke quality ratings | correlation coefficient: 0.4918 -> 0.5510 | Improved |
Closest papers by meaning
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · similarity 0.84)
- Learning Interpretable Hierarchical Dynamical Systems Models from Time Series Data (ICLR 2025 · similarity 0.84)
- A scalable generative model for dynamical system reconstruction from neuroimaging data (NeurIPS 2024 · similarity 0.83)
- Almost-Linear RNNs Yield Highly Interpretable Symbolic Codes in Dynamical Systems Reconstruction (NeurIPS 2024 · similarity 0.83)
- Learning Physics From Video: Unsupervised Physical Parameter Estimation for Continuous Dynamical Systems (CVPR 2025 · similarity 0.83)
- Meta-Dynamical State Space Models for Integrative Neural Data Analysis (ICLR 2025 · similarity 0.83)
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.
