Research map / Probabilistic and scientific modeling / Variational and generative sampling
Exploiting locality in high-dimensional Factorial hidden Markov models
Lorenzo Rimella, Nick Whiteley
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 Graph Filter-Smoother (localized Bayes update) instead of exact-cost methods for high-dimensional FHMM inference | London Underground FHMM (M=48) · London Underground passenger flow | RMSE no missing data: 4.796 vs LSTM 4.639 | Mixed |
| Graph Filter-Smoother in place of LSTM for prediction with missing data | London Underground passenger flow · Missing data windows | RMSE and its variability across runs: 5.399 ± 0.014 vs 7.427 ± 2.080 (quiet); 5.493 ± 0.019 vs 6.178 ± 1.425 (peak) | Improved |
| Increase localization radius m from 0 to 1 in Graph Smoother EM (ablation) | Synthetic FHMM M=3 parameter estimation · Synthetic FHMM M=3 | estimate accuracy of c and sigma^2: slight increase in accuracy | Improved |
| Increase localization radius m from 0 to 1 for transition probability estimates (ablation) | Synthetic FHMM M=10 parameter estimation · Synthetic FHMM M=10 | accuracy of transition probabilities: m=0 slightly more accurate than m=1 | Hurt |
| Replace Graph Smoother EM with fully (completely) decoupled variational approximation | Synthetic FHMM parameter estimation · Synthetic FHMM M=3 and M=10 | parameter estimate accuracy: completely decoupled VB transition estimates poor (e.g., p(1,0)=0.084 vs true 0.200 at M=3) | Hurt |
| Increase localization radius m in Graph Filter/Smoother (ablation) | Synthetic FHMM chain (M up to 10) · Smoothing LTV error | local total variation error | Improved |
| Increase number of chain dimensions M with fixed m (ablation) | Graph Filter-Smoother on synthetic chain FHMM · Synthetic FHMM smoothing LTV | average LTV error | No clear effect |
Closest papers by meaning
- Scaling Factorial Hidden Markov Models: Stochastic Variational Inference without Messages (NeurIPS 2016 · similarity 0.87)
- Sparse Markov Models for High-dimensional Inference (JMLR 2023 · similarity 0.86)
- Topological Hidden Markov Models (NeurIPS 2024 · similarity 0.85)
- A Spectral Algorithm for Inference in Hidden semi-Markov Models (JMLR 2017 · similarity 0.83)
- An Infinite Hidden Markov Model With Similarity-Biased Transitions (ICML 2017 · similarity 0.83)
- Pre-trained Large Language Models Learn to Predict Hidden Markov Models In-context (NeurIPS 2025 · similarity 0.81)
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.
