Accelerating Sustainable Glass Discovery: Integrating Molecular Dynamics, Machine Learning, and Robotic Synthesis

This Zenodo archive contains datasets, trained models, and supporting code associated with the paper “Accelerating Sustainable Glass Discovery: Integrating Molecular Dynamics, Machine Learning, and Robotic Synthesis”. Contents mlregression: The mlregression package provides wrappers around PyTorch and GPyTorch classes to enable a workflow compatible with scikit-learn (e.g., unified fit/predict interfaces and pipeline-style usage). glass_sim_helper: The glass_sim_helper package contains helper functions for generating input files required to run high-throughput molecular dynamics (HT-MD) glass simulations. Example usages are provided in the folder example_glass_sim. Environment requirements (*.txt): The corresponding Python environment requirements are provided in the respective .txt files (one file per environment). models_data_V14.zip and models_data_V18.zip: These ZIP archives contain the datasets used in this work as well as stored (trained) models. The models and datasets can be created, loaded, and processed using the mlregression package. iterative_learning: This folder contains scripts for running the iterative learning workflow. Notes The included requirements files define the dependencies needed to reproduce the software environments used for training and model handling. The datasets and stored models are organized by dataset version (V14, V18) and are provided as ZIP archives.