Giters

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

Repository from Github https://github.comaxsk/augmented-jump-chainRepository from Github https://github.comaxsk/augmented-jump-chain

ajcisokannpythonjuliascrapbooksqracommittorneural-networkoptimizationzibpaper

augmented-jump-chain (former name birthdeath)

Amongst the rest, most prominently features the code for the The Augmented Jump Chain (see 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

About

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

ajcisokannpythonjuliascrapbooksqracommittorneural-networkoptimizationzibpaper

Languages

Language:Jupyter Notebook 99.3%Language:Python 0.4%Language:Julia 0.3%Language:MATLAB 0.1%Language:M 0.0%