AL4ECG : Active Learning for Electronic Coarse Graining

Documentation for the active learning (AL) workflow developed as a part of the article Sivaraman, G.; Jackson, N.E. "Coarse-Grained Density Functional Theory Predictions via Deep Kernel Learning" J. Chem. Theory Comput. 2022, 18, 2, 1129-1141. For more details, please refer to the paper.

What are the capabilities of AL4ECG workflow ?

The workflow is built as a PyTorch based GPU accelerated framework and offers the following capabilities:

• GPU accelerated Scalable Gaussian Processes and Exact Deep Kernel Learning (DKL) based on GPyTorch library
• Bayesian Optimization for hyperparameter tuning of DKL using the GPyOpt library
• PyTorch based numeric implementation of AL query strategies using standard GPR based uncertainty and beyond.
• Capable of running on the state-of-the-art NVIDIA A100 GPU's available at the LCRC SWING cluster and the ALCF ThetaGPU.

What are the type of supervised learning method and AL queries are supported in the AL4ECG workflow?

Supervised learning methods

• Gaussian Process Regression: Single and Multi-task
• Deep Kernel Learning

Kernels

• Matern
• RBF

AL queries

• Random
• Uncertainty
• Expected Model Output Changes (EMOC)

Installation

Running GPyTorch on A100 GPU has the following basic requirment:

• MAGMA + CUDA 11.0

The step-by-step compilation is covered in INSTALLATION.MD

How to run the workflow?

Running the workflow is extensively covered in RUN.MD

How do i cite AL4ECG workflow ?

If you are using this active learning workflow in your research paper, please cite us as

@article{doi:10.1021/acs.jctc.1c01001,
author = {Sivaraman, Ganesh and Jackson, Nicholas E.},
title = {Coarse-Grained Density Functional Theory Predictions via Deep Kernel Learning},
journal = {Journal of Chemical Theory and Computation},
volume = {18},
number = {2},
pages = {1129-1141},
year = {2022},
doi = {10.1021/acs.jctc.1c01001},
    note ={PMID: 35020388},

URL = { 
        https://doi.org/10.1021/acs.jctc.1c01001
    
},
eprint = { 
        https://doi.org/10.1021/acs.jctc.1c01001
    
}

}

Acknowledgements

This material is based upon work supported by Laboratory Directed Research and Development (LDRD-CLS-1-630) funding from Argonne National Laboratory, provided by the Director, Office of Science, of the U.S. Department of Energy under Contract No. DE-AC02-06CH11357. N.E.J acknowledges support from the Dreyfus Program for Machine Learning in the Chemical Sciences and Engineering during this project. This research used resources of the Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC02-06CH11357. Argonne National Laboratory’s work was supported by the U.S. Department of Energy, Office of Science, under contract DE-AC02-06CH11357. We gratefully acknowledge the computing resources provided on Bebop and Swing, high-performance computing clusters operated by the Laboratory Computing Resource Center at Argonne National Laboratory.

Code contributor

• Ganesh Sivaraman gsivaraman:AT::anl.gov
• Nicholas Jackson jackson.nick.e::aT:gmail.com