Sovereign 8s.is mirror of standardgalactic/MatterVial_SupportData
  • Python 97.8%
  • Shell 2.2%
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2025-11-08 20:52:44 +01:00
.github Updated with results on MatBench 2025-09-12 17:42:31 +02:00
mattervial_perovskites_full_analysis update readmes 2025-09-15 16:11:02 +02:00
modnet_mattervial_matbench script to download dataset 2025-09-15 16:15:19 +02:00
other_datasets New folder with additional datasets to validate MatterVial included 2025-11-08 20:52:44 +01:00
results_encoders_and_SI_data finalized updating the support data 2025-09-15 15:18:01 +02:00
.gitignore out files added 2025-09-12 18:32:51 +02:00
README.md update readmes 2025-09-15 16:11:02 +02:00

MatterVial Support Data

This repository contains the data and code used to support the findings in the paper:

"Combining feature-based approaches with graph neural networks and symbolic regression for synergistic performance and interpretability"
by Rogério Almeida Gouvêa, Pierre-Paul De Breuck, Tatiane Pretto, Gian-Marco Rignanese, Marcos José Leite Santos

Abstract

This study introduces MatterVial, an innovative hybrid framework for feature-based machine learning in materials science. MatterVial expands the feature space by integrating latent representations from a diverse suite of pretrained graph neural network (GNN) models—including structure-based (MEGNet), composition-based (ROOST), and equivariant (ORB) graph networks—with computationally efficient, GNN-approximated descriptors and novel features from symbolic regression. Our approach combines the chemical transparency of traditional feature-based models with the predictive power of deep learning architectures. When augmenting the feature-based model MODNet on Matbench tasks, this method yields significant error reductions and elevates its performance to be competitive with, and in several cases superior to, state-of-the-art end-to-end GNNs, with accuracy increases exceeding 40% for multiple tasks. An integrated interpretability module, employing surrogate models and symbolic regression, decodes the latent GNN-derived descriptors into explicit, physically meaningful formulas. This unified framework advances materials informatics by providing a high-performance, transparent tool that aligns with the principles of explainable AI, paving the way for more targeted and autonomous materials discovery.

Repository Contents

This repository includes:

  • Results for the autoencoder and previous version of MatterVial (pGNN): The detailed results testing the different descriptor-oriented featurizers (l-MM and l-OFM)and all results obtained from the previously less general package which are now on the SI of the paper.

    • results_encoders_and_SI_data/
  • Featurized datasets and models: Featurized datasets and final models for the main results of the paper. These are downloaded from FigShare link, a python script is provided to help with this.

  • download_datasets_and_models.py

  • Matbench v0.1 evaluation of MODNet@MatterVial models: All general results that are shown in Table 1 of the paper are shown here including output files and fold by fold metrics.

    • modnet_mattervial_matbench_results/ Correspond to Table 1 of the paper:

benchmark results

Results table 2, focusing on matbench_perovskites dataset:

  • MODNet@MatterVial with adjacent models:
    • mattervial_perovskites_full_analysis

benchmark results

Getting Started

To replicate the results presented in the paper or to apply these methods to your own dataset, check MatterVial repository enhanced with the GNN featurizers, and check the scripts provided along with the results.

Contact and Contributions

For any clarifications, please contact the corresponding authors: