A Recipe for Charge Density Prediction
Authors: Xiang Fu, Andrew Rosen, Kyle Bystrom, Rui Wang, Albert Musaelian, Boris Kozinsky, Tess Smidt, Tommi Jaakkola
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We apply our recipe to the widely used QM9 charge density benchmark [18, 19, 7]. Our method outperforms existing state-of-the-art methods while being around 30% faster. |
| Researcher Affiliation | Collaboration | Xiang Fu1 Andrew Rosen2,3 Kyle Bystrom4 Rui Wang1 Albert Musaelian4 Boris Kozinsky4,5 Tess Smidt1 Tommi Jaakkola1 1Massachusetts Institute of Technology 2UC Berkeley 3Lawrence Berkeley National Laboratory 4Harvard John A. Paulson School Of Engineering and Applied Sciences 5Robert Bosch Research and Technology Center |
| Pseudocode | Yes | Algorithm 1 Pseudo code for the charge density prediction procedure |
| Open Source Code | Yes | Code is available at https://github.com/kyonofx/scdp. |
| Open Datasets | Yes | The QM9 charge density dataset [18, 19, 7] contains charge density calculations for 133,845 small organic molecules using the Vienna Ab initio Simulation Package (VASP). We adopt the original split, where 123,835, 50, and 10,000 data points are used for training, validation, and testing, respectively. [...] The data used in this paper is publicly available [7]. |
| Dataset Splits | Yes | We adopt the original split, where 123,835, 50, and 10,000 data points are used for training, validation, and testing, respectively. |
| Hardware Specification | Yes | We benchmark the efficiency of different methods by the number of molecules predicted per minute (Mol. per min.) on a single NVIDIA A100-80GB-PCIe GPU for the QM9 test split. |
| Software Dependencies | Yes | Basis-set-exchange-v0.9.1 [36, 37] and Py SCF-v2.5.0 [39] are used to build the orbital basis sets. E3NN-v0.5.1 [28], Py Torch-v1.13.1 [59], and CUDA-v11.6 [60] are used to build the SCDP models. |
| Experiment Setup | Yes | Table 3: Hyperparameters for SCDP models on the QM9 dataset. |