Effect of t_stride on committor estimation

Determine the nature and magnitude of the effect of the stride-time parameter t_stride on committor estimation when swarms are evolved in multiples of t_stride and basin membership is checked only at these intervals in the proposed importance-sampling procedure for learning the committor function.

Background

To avoid issues associated with very small lag times τ during swarm-based sampling, the authors introduce a stride parameter t_stride and evolve each swarm in multiples of this interval before checking for a stopping criterion. This creates an effective minimum τ and may affect the learned committor near metastable basins.

Although the authors observe that a larger t_stride may theoretically bias committor estimation near basins, they report that such bias has minimal effect on rate estimation in their experiments. However, they explicitly leave a rigorous characterization of how t_stride influences committor estimation itself as future work.