Hybrid Directional Graph Neural Network for Molecules

Authors: Junyi An, Chao Qu, Zhipeng Zhou, Fenglei Cao, Xu Yinghui, Yuan Qi, Furao Shen

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the performance of HDGNN on the QM9 dataset and the IS2RE dataset of OC20, demonstrating its state-of-the-art performance on several tasks and competitive performance on others. Our code is anonymously released on https://github.com/ajy112/HDGNN. ... We provide extensive ablation studies on non-equivariant modules and the network structure. ... In this section, we conduct experiments to investigate the effectiveness of proposed method over Quantum Machines 9 (QM9) (Ramakrishnan et al., 2014) and IS2RE task in Open Catalyst 2020 (OC20) (Chanussot et al., 2021) benchmarks.
Researcher Affiliation Collaboration Junyi An1 , Chao Qu2 , Zhipeng Zhou2, Fenglei Cao3, Yinghui Xu4, Yuan Qi4, Furao Shen1 1State Key Laboratory for Novel Software Technology, Nanjing University 2INFLY TECH (Shanghai) Co., Ltd. 3Shanghai Academy of Artificial Intelligence for Science 4Artificial Intelligence Innovation and Incubation (AI3) Institute, Fudan University
Pseudocode No The paper describes algorithms and methods in text and uses diagrams, but does not include explicit pseudocode blocks or formally labeled 'Algorithm' sections.
Open Source Code Yes Our code is anonymously released on https://github.com/ajy112/HDGNN.
Open Datasets Yes We evaluate our HDGNN on the QM9 benchmark (Ramakrishnan et al., 2014) and the IS2RE dataset of OC20 (Chanussot et al., 2021). The OC20 dataset contains over 130 million structures used to train models for predicting forces and energies during structure relaxations with a CC Attribution 4.0 License.
Dataset Splits No The paper mentions training and test sets and splitting the test set, but it does not explicitly provide percentages or specific counts for a validation set, nor does it specify predefined validation splits with citations.
Hardware Specification No The paper does not provide specific hardware details such as GPU models, CPU specifications, or memory used for running the experiments.
Software Dependencies No The paper mentions using a '2-layer MLP with the Si LU activation function' but does not specify software dependencies like libraries or frameworks with version numbers (e.g., PyTorch 1.9, TensorFlow 2.x).
Experiment Setup Yes The baseline denotes a HDGNN where K = 8, C = 64 and L=6. ... In our implementation, we utilize a 2-layer MLP with the Si LU activation function.