/Bayes-ML-sld

Primary LanguageJupyter NotebookBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Bayes ML model to predict SLD distribution

Create the environment:

conda env create -f play_env.yml
source activate playground

Create a training set:

python train.py  -n 100000 -v 1000 -f config-erik.json --create

Train the ML model:

python train.py  -n 100000 -v 1000 -f config-erik.json

Note: See the following example to call a fit from code insteand of using the refl1d client: https://github.com/pozzo-research-group/c-wolf-blends-morphology/blob/master/sans/PS_Fitting/Polystyrene_Fits.ipynb

TODO:

  • Add more layers
  • Get rid of post-processing to get proper sld
  • Add full SLD to reconstruction loss
  • Add q resolution
  • Check max thickness and get rid of model and generate another one.