search

mertyigit / INDEEDopt

An optimization framework using Latin Hypercube Design and deep learning to explore high-dimensional parameter spaces.
MIT License
26 stars 10 forks source link

readme

INDEEDopt: INitial DEsign Enhanced Deep learning OPTimization framework for high dimensional ReaxFF force field parameterization

INDEEDopt Workflow Scheme

Python PyTorch GitHub License

UPDATES

Due to time limitations in other projects, I won't be able to maintain this package.

A common error that users of INDEEDopt face is the generation of lousy ffield files. The reason for that is the extra spaces in ffield file that are located in the "folder". Please clean all extra spaces after each line in ffield file.

Added an example INDEEDopt folder named example_github.zip. (November 2021)

Recent updates to some subroutines due to Tensorflow Deep Learning library updates (Feb 2021)