FAENet: Frame Averaging Equivariant GNN for Materials Modeling
Authors: Alexandre Agm Duval, Victor Schmidt, Alex Hernández-Garcı́a, Santiago Miret, Fragkiskos D. Malliaros, Yoshua Bengio, David Rolnick
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We prove the validity of our method theoretically and empirically demonstrate its superior accuracy and computational scalability in materials modeling on the OC20 dataset (S2EF, IS2RE) as well as common molecular modeling tasks (QM9, QM7-X). and 5. Experiments. |
| Researcher Affiliation | Collaboration | 1Universit e Paris-Saclay, Centrale Sup elec, Inria 2Mila Quebec AI Institute 3Intel Labs 4Universit e de Montr eal 5Mc Gill Unversity. |
| Pseudocode | No | The paper includes architectural diagrams and descriptions of its components (e.g., Figure 1 'Overview of FAENet architecture'), but it does not present structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | A package implementation is available at https: //faenet.readthedocs.io. |
| Open Datasets | Yes | OC20 (Zitnick et al., 2020) is a large dataset for catalysis discovery. and QM9 (Ramakrishnan et al., 2014) is a widely used dataset for molecular property prediction. and QM7-X (Hoja et al., 2021) is a dataset containing 7K molecular graphs... |
| Dataset Splits | Yes | It comes with a predefined train/val/test split, 450,000 training samples and hidden test labels. and The dataset is split into 110K molecules for training, 10K for validation, and remaining 14K for testing. |
| Hardware Specification | Yes | Experiments were run on 1 NVIDIA RTX8000 GPUs. |
| Software Dependencies | Yes | Pytorch v.1.13 (Paszke et al., 2019), Py Torch Geometric v2.2.0 (Fey & Lenssen, 2019) Num Py v1.23.5 (Harris et al., 2020) |
| Experiment Setup | Yes | We detail FAENet s list of hyperparameters for all four datasets IS2RE, S2EF, QM7-X and QM9 in Table 7. |