A Group Symmetric Stochastic Differential Equation Model for Molecule Multi-modal Pretraining

Authors: Shengchao Liu, Weitao Du, Zhi-Ming Ma, Hongyu Guo, Jian Tang

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental By comparing with 17 pretraining baselines, we empirically verify that Molecule SDE can learn an expressive representation with state-of-the-art performance on 26 out of 32 downstream tasks.
Researcher Affiliation Academia 1Mila Quebec AI Institute, Canada 2Universite de Montreal, Canada 3Chinese Academy of Sciences, China 4National Research Council Canada, Canada 5HEC Montreal, Canada.
Pseudocode No The paper includes detailed pipeline diagrams (e.g., Figure 3, Figure 4) but does not present any formal pseudocode or algorithm blocks.
Open Source Code Yes The source codes are available in this repository.
Open Datasets Yes For pretraining, we use PCQM4Mv2 (Hu et al., 2020a). ... We take 110K for training, 10K for validation, and 11K for testing.
Dataset Splits Yes We take 110K for training, 10K for validation, and 11K for testing.
Hardware Specification Yes All the jobs are running on one single V100 GPU card.
Software Dependencies No The paper mentions software components like GIN and Sch Net models but does not provide specific version numbers for these or other ancillary software dependencies (e.g., Python, PyTorch, CUDA versions).
Experiment Setup Yes Table 8. Hyperparameter specifications for Molecule SDE. Hyperparameter Value epochs {50, 100} learning rate 2D GNN {1e-5, 1e-6} SDE option {VE, VP} masking ratio M {0, 0.3} number of steps {1000}.