Molecular Property Prediction: A Multilevel Quantum Interactions Modeling Perspective
Authors: Chengqiang Lu, Qi Liu, Chao Wang, Zhenya Huang, Peize Lin, Lixin He1052-1060
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on both datasets of equilibrium and off-equilibrium molecules demonstrate the effectiveness of our model. Furthermore, the detailed results also prove that MGCN is generalizable and transferable for the prediction. |
| Researcher Affiliation | Academia | Anhui Province Key Lab. of Big Data Analysis and Application, University of S&T of China Key Laboratory of Quantum Information, University of S&T of China {qiliuql, helx}@ustc.edu.cn, {lunar, wdyx2012, huangzhy, linpz}@mail.ustc.edu.cn |
| Pseudocode | No | The paper describes the model architecture and mathematical formulations but does not include a separate section or figure explicitly labeled as "Pseudocode" or "Algorithm." |
| Open Source Code | No | The paper does not contain any explicit statement about releasing the source code for the methodology or provide a link to a code repository. |
| Open Datasets | Yes | QM9. The QM91 dataset (Ramakrishnan et al. 2014) is perhaps the most well-known benchmark dataset which contains 134k equilibrium molecules... http://www.quantum-machine.org/datasets/#qm9 ANI-1. The ANI-12 dataset provides access to the total energies of 20 million off-equilibrium molecules... https://www.nature.com/articles/sdata2017193 |
| Dataset Splits | Yes | For all 13 properties of QM9, we pick 110k out of 130k molecules randomly as our training set that accounts for about 84.7% of the entire dataset. With the rest of the data, we choose half of them as the validation set and the other half as the testing set. As for the much larger ANI-1, we randomly choose 90% samples for training, 5% samples for validation and 5% for testing. |
| Hardware Specification | Yes | Experimentally, with the same setting (a single core of a Xeon E5-2660), our model spends 2.4 10 2 second predicting the property of one molecule, which is nearly 1.5 105 times faster than DFT. |
| Software Dependencies | No | The paper mentions "Adam optimizer (Kingma and Ba 2014)" but does not provide specific version numbers for any key software components or libraries like Python, PyTorch, or TensorFlow. |
| Experiment Setup | Yes | We use mini-batch stochastic gradient descent (mini-batch SGD) with the Adam optimizer (Kingma and Ba 2014) to train our MGCN. The batch size is set to 64 and the initial learning rate is 1e 5. |