GemNet: Universal Directional Graph Neural Networks for Molecules

Authors: Johannes Gasteiger, Florian Becker, Stephan Günnemann

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the benefits of the proposed changes in multiple ablation studies. Gem Net outperforms previous models on the COLL, MD17, and OC20 datasets by 34 %, 41 %, and 20 %, respectively, and performs especially well on the most challenging molecules.
Researcher Affiliation Academia Johannes Gasteiger, Florian Becker, Stephan Günnemann Technical University of Munich, Germany {j.gasteiger,beckerf,guennemann}@in.tum.de
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Our implementation is available online. https://www.daml.in.tum.de/gemnet
Open Datasets Yes We evaluate our model on four molecular dynamics datasets. COLL [28, CC-BY 4.0] consists of configurations taken from molecular collisions of different small organic molecules. MD17 [9] consists of configurations of multiple separate, thermalized molecules, considering only one molecule at a time. MD17@CCSD [10] uses the same setup, but calculates the forces using the more accurate and expensive CCSD or CCSD(T) method. The open catalyst (OC20) dataset [8, CC-BY 4.0] consists of energy relaxation trajectories of solid catalysts with adsorbate molecules.
Dataset Splits Yes Following the setup of Batzner et al. [4], we use 1000 training and validation configurations for MD17, and 950 training and 50 validation configurations for MD17@CCSD.
Hardware Specification No No specific hardware details (such as GPU/CPU models, memory specifications, or cluster configurations) used for running experiments were provided in the paper.
Software Dependencies No The paper does not provide specific version numbers for ancillary software components (e.g., Python, PyTorch, CUDA, specific libraries or solvers).
Experiment Setup Yes We train our models using the AdamW optimizer [40] with a learning rate of 1e-4, an exponential decay of 0.01 per 100 000 steps, a batch size of 32, and an extensive search for the optimal force loss weight (which varied across datasets).