All reweighted paths lead to Rome: Unification of reweighting schemes for stratified path sampling
Abstract
Path sampling-based stratification algorithms such as Transition Interface Sampling (TIS) enable the evaluation of kinetic rate constants and mechanisms of complex molecular rare event transitions. By sampling only short dynamical trajectories, TIS allows constructing an estimate of the unbiased dynamics by reweighting the interface path ensembles. In the past decades several methods have been developed and applied to achieve such path reweighting, based on statistically optimal use of the trajectory data. Why and how this reweighting exactly works has been underexplored up to now. Here we discuss several different schemes to compute the reweighted path ensemble (RPE), notably point-wise matching, Weighted Histogram Analysis Method (WHAM), Multistate Bennett Acceptance Ratio (MBAR), variance based reweighting (VBR) and survival analysis reweighting (SAR). We show that except for the point-wise matching the other methods are leading to the same outcome, and in fact can be seen as identical. Moreover, the SAR links to a well-established nonparametric maximum-likelihood estimator of product-limit form. This allows a synthesis of all path reweighting schemes, identifying the common underlying nature. All schemes follow from this common origin. We illustrate the reweighting with some simple examples, and show that care has to be taken when applying WHAM, if the number of interfaces and histogram bins do not match.
Related
- Robustness of oxygen metrics to sampling frequency in a stratified lake.
- Importance-Reweighted Fock-Space Variational Monte Carlo
- A Green-function unification of weak interactions with environmental dressing
- Efficient measurement schemes for the Monte Carlo projective quantum eigensolver
- Assessing the performance of coupled-trajectory schemes on full-dimensional two-state linear vibronic coupling models
BibTeX
@misc{bolhuis2026reweighted,
title = {{All reweighted paths lead to Rome: Unification of reweighting schemes for stratified path sampling}},
author = {Peter G. Bolhuis and Rik S. Breebaart and Gianmarco Lazzeri and Jan-Hendrik Prinz and Roberto Covino and Titus S. van Erp},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.06283},
eprint = {2611.06283},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.06283},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - All reweighted paths lead to Rome: Unification of reweighting schemes for stratified path sampling
AU - Peter G. Bolhuis
AU - Rik S. Breebaart
AU - Gianmarco Lazzeri
AU - Jan-Hendrik Prinz
AU - Roberto Covino
AU - Titus S. van Erp
PY - 2026
DA - 2026/11/04
PB - arXiv
DO - 10.48550/arxiv.2611.06283
UR - https://arxiv.org/abs/2611.06283
AB - Path sampling-based stratification algorithms such as Transition Interface Sampling (TIS) enable the evaluation of kinetic rate constants and mechanisms of complex molecular rare event transitions. By sampling only short dynamical trajectories, TIS allows constructing an estimate of the unbiased dynamics by reweighting the interface path ensembles. In the past decades several methods have been developed and applied to achieve such path reweighting, based on statistically optimal use of the trajectory data. Why and how this reweighting exactly works has been underexplored up to now. Here we discuss several different schemes to compute the reweighted path ensemble (RPE), notably point-wise matching, Weighted Histogram Analysis Method (WHAM), Multistate Bennett Acceptance Ratio (MBAR), variance based reweighting (VBR) and survival analysis reweighting (SAR). We show that except for the point-wise matching the other methods are leading to the same outcome, and in fact can be seen as identical. Moreover, the SAR links to a well-established nonparametric maximum-likelihood estimator of product-limit form. This allows a synthesis of all path reweighting schemes, identifying the common underlying nature. All schemes follow from this common origin. We illustrate the reweighting with some simple examples, and show that care has to be taken when applying WHAM, if the number of interfaces and histogram bins do not match.
ER -