Research map / Learning theory and algorithms / Optimal transport
A Distance for HMMs Based on Aggregated Wasserstein Metric and State Registration
Yukun Chen, Jianbo Ye, Jia Li
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 MAW instead of KL divergence as HMM distance | Adaboost motion classifier on Mocap · Mocap (6 joint groups) | test accuracy (%): 73.37% -> 85.21% | Improved |
| use IAW instead of KL divergence as HMM distance | Adaboost motion classifier on Mocap · Mocap (6 joint groups) | test accuracy (%): 73.37% -> 88.17% | Improved |
| use MAW instead of KL divergence as HMM distance | Adaboost motion classifier on Mocap · Mocap (27 joint groups) | test accuracy (%): 88.76% -> 95.27% | Improved |
| use IAW instead of KL divergence as HMM distance | Adaboost motion classifier on Mocap · Mocap (27 joint groups) | test accuracy (%): 88.76% -> 94.08% | Improved |
| use MAW/IAW instead of KL for retrieval under perturbed means and transition matrices | GMM-HMM retrieval on synthetic data · Synthetic perturbation of means and transitions | precision-recall | Improved |
| use MAW/IAW instead of KL for retrieval under perturbed covariances | GMM-HMM retrieval on synthetic data · Synthetic perturbation of covariances | precision-recall | Hurt |
| add state registration (IAW) to MAW | MAW · Synthetic covariance perturbation retrieval | precision-recall | Improved |
| use MAW/IAW instead of KL for motion retrieval | Mocap per-joint-group retrieval · Mocap retrieval | precision-recall | Improved |
| use IAW instead of MAW | Mocap classification · Mocap 6 vs 27 joint groups | test accuracy (%): 85.21% -> 88.17% (6 joints); 95.27% -> 94.08% (27 joints) | Mixed |
| use Wasserstein distance instead of KL | distance between estimated Gaussians · Toy experiment, mean shift | distance estimation variance | Improved |
Closest papers by meaning
- Aggregated Wasserstein Distance and State Registration for Hidden Markov Models (TPAMI 2019 · similarity 0.99)
- Order-Preserving Wasserstein Distance for Sequence Matching (CVPR 2017 · similarity 0.86)
- FlowHMM: Flow-based continuous hidden Markov models (NeurIPS 2022 · similarity 0.85)
- Frontiers to the learning of nonparametric hidden Markov models (JMLR 2025 · similarity 0.85)
- A New Robust Partial p-Wasserstein-Based Metric for Comparing Distributions (ICML 2024 · similarity 0.85)
- An Infinite Hidden Markov Model With Similarity-Biased Transitions (ICML 2017 · 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.
