Adaptive-MFML

Scripts for Adaptive sampling for MFML. Includes scripts for plotting and hyper-parameter optimization. Datasets used are open sourced and can be accessed at their respective repositories.
This code repository accompanies the preprint at [TBA].

The files are as follows:

1. ANI_preprocessing.py extracts ANI data, samples 50k random geometries along with their multifidelity energies. It also generates the SLATM descriptor for this dataset.
2. MFMLmanager.py contains the main bulk of the sampling strategies. It performs training and prediction with the MFML model.
3. MFML_Model.py is the script that runs the MFML scheme given a multifidelity training dataset.
4. basic_experiments.py contains scripts to run single fidelity and basic MFML learning curves.
5. avg_experiments.py runs the collection of different experiments from the manuscript. The appropriate line should be uncommented.
6. plottinghelpers.py contains functions to assist with generating the plots.
7. Plots.ipynb is the jupyter notebook with the plots.
8. requirements.txt is the list of python packages used.
9. hyperopt.py carries out hyper-parameter optimization for kernel parameters using a GP-based Bayesian minimizer.
10. dataloaders.py is a helper function to load datasets.

The datasets used are all open sources and should be downloaded before running these scripts.