https://github.com/axsk/optimpsampling.jl

Optimal Importance Sampling for Diffusion processes applied to ISOKANN.
https://github.com/axsk/optimpsampling.jl

importance-sampling isokann machine-learning optimal-control reinforce scrapbook zib

Last synced: 7 months ago
JSON representation

Optimal Importance Sampling for Diffusion processes applied to ISOKANN.

• Host: GitHub
• URL: https://github.com/axsk/optimpsampling.jl
• Owner: axsk
• Created: 2022-02-24T14:03:54.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2023-01-05T11:12:34.000Z (almost 3 years ago)
• Last Synced: 2024-03-24T06:10:17.129Z (over 1 year ago)
• Topics: importance-sampling, isokann, machine-learning, optimal-control, reinforce, scrapbook, zib
• Language: Julia
• Homepage:
• Size: 1.23 MB
• Stars: 0
• Watchers: 3
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# OptImpSampling.jl

A WIP collection of algorithms related to optimal importance sampling (OIS).

With a focus on ISOKANN:

- [ociso](src/ociso.jl): OIS for Koopman eigen- and chi functions

- [isokann](src/isokann.jl): ISOKANN using OIS for the chi approximations

- [control](src/control.jl): simpler rewrite of ociso

- [isonew](src/isonew.jl): rewrite of isokann, uses humboldtsampling

- [isokann2](src/isokann2.jl): sketch of ISOKANN in higher dimensions

Utils:

- [sqra](src/sqra.jl): Square root approximation for the generators of Ito diffusions (copy from Sqra.jl)

- [humboldtsample](src/humboldtsample.jl): sample uniform along the reaction coordinate

Experimental implementations:

- [logvar](src/logvar.jl): OIS  of path functionals with control variates (implements Richter, Nusken 2021)

- [reinforce](src/reinforce.jl): OIS of functions via REINFORCE

- [reinforcepath](src/reinforcepath.jl): like reinfore but on path space (equivalent to logvar)

- [fbsde](src/fbsde.jl): Forward backward SDE with Neural Controls, learning all times at once (like Kebiri, Hartmann 2019)