https://github.com/axsk/augmented-jump-chain
Continuation of AJC work (2020-21) with trips to hitting time optimization, spa, sqra, isokann, committor neural network, sparse boxes, voronoi linear program, meta sgd, adaptive euler maruyama
https://github.com/axsk/augmented-jump-chain
ajc committor isokann julia neural-network optimization paper python scrapbook sqra zib
Last synced: about 1 month ago
JSON representation
Continuation of AJC work (2020-21) with trips to hitting time optimization, spa, sqra, isokann, committor neural network, sparse boxes, voronoi linear program, meta sgd, adaptive euler maruyama
- Host: GitHub
- URL: https://github.com/axsk/augmented-jump-chain
- Owner: axsk
- Created: 2020-08-13T12:58:28.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2023-08-29T15:19:48.000Z (almost 3 years ago)
- Last Synced: 2025-06-04T04:10:44.215Z (about 1 year ago)
- Topics: ajc, committor, isokann, julia, neural-network, optimization, paper, python, scrapbook, sqra, zib
- Language: Jupyter Notebook
- Homepage:
- Size: 15.4 MB
- Stars: 0
- Watchers: 3
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# augmented-jump-chain (former name `birthdeath`)
Amongst the rest, most prominently features the code for the [The Augmented Jump Chain](https://doi.org/10.1002/adts.202000274) (see [src/ajcs.py](src/ajcs.py))
## Contents
### Part 1 (python)
- **EAMC** + Paper Plots
- Hitting times + Optimization (ADAM, RProp, Momentum, RMSProp)
- Adjoint ODE solver as in SUNDIALS (for optimization of hitting times?)
- Temporal Gillespie
- SPA
- SQRA (ndtorus, perturbation, derivatives for the adjoint problem)
### Part 2 (julia) (most is now in Sqra.jl)
- ISOKANN experiments
- Committor neural network
- Autodiff Bug MWEs
- SparseBoxes, Sqra, picking
- voronoi neighborhood by linear program of H. Lie
- meta SGD
- adaptive euler maruyama
# History
WIP from 07.20 to 06.21
Continuation of https://github.com/axsk/generators and in a similar dirty state
Most usefull ideas where ported to ttps://github.com/axsk/Sqra.jl