Efficient and Equivariant Graph Networks for Predicting Quantum Hamiltonian
Authors: Haiyang Yu, Zhao Xu, Xiaofeng Qian, Xiaoning Qian, Shuiwang Ji
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We perform experiments on MD17 datasets, including four molecular systems. Experimental results show that our QHNet can achieve comparable performance to the state of the art methods at a significantly faster speed. |
| Researcher Affiliation | Academia | 1Department of Computer Science & Engineering, Texas A&M University, TX, USA 2Department of Materials Science & Engineering, Texas A&M University, TX, USA 3Department of Electrical & Computer Engineering, Texas A&M University, TX, USA 4Department of Physics & Astronomy, Texas A&M University, TX, USA. |
| Pseudocode | No | The paper does not contain any pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | Yes | Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS). |
| Open Datasets | Yes | We conduct experiments to evaluate the performance of QHNet on MD17 datasets (Sch utt et al., 2019). The statistics of MD17 dataset is shown in Table 5. |
| Dataset Splits | Yes | Table 5. The statistics of MD17 dataset (Sch utt et al., 2019). Dataset # of structures Train Val Test ... Water 4,900 500 500 3,900 |
| Hardware Specification | Yes | In our experiments, models are trained on a single 11GB Nvidia Ge Force RTX 2080Ti GPU and Intel Xeon Gold 6248 CPU. |
| Software Dependencies | Yes | Our experiments are implemented based on Py Torch 1.11.0 (Paszke et al., 2019), Py Torch Geometric 2.1.0 (Fey & Lenssen, 2019), and e3nn (Geiger & Smidt, 2022). |
| Experiment Setup | Yes | Specifically, the scheduler increases the learning rate gradually during the first 1,000 warm-up steps. The initial learning rate is 0, and the maximum learning rate is 5e 4. Then, the scheduler reduces the learning rate linearly so that the learning rate reaches 1e 7 at the last step. ... for QHNet, the batch size is set to 10 for water, ethanol, and malondialdehyde, while it is set to 5 for uracil. |