Conditional Graph Information Bottleneck for Molecular Relational Learning

Authors: Namkyeong Lee, Dongmin Hyun, Gyoung S. Na, Sungwon Kim, Junseok Lee, Chanyoung Park

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on various tasks with real-world datasets demonstrate the superiority of CGIB over state-of-the-art baselines.
Researcher Affiliation Academia 1KAIST 2POSTECH 3KRICT. Correspondence to: Chanyoung Park <cy.park@kaist.ac.kr>.
Pseudocode No The paper describes its methodology using text and mathematical equations in Section 4, but it does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Our code is available at https: //github.com/Namkyeong/CGIB.
Open Datasets Yes For the molecular interaction prediction task, we use Chromophore dataset (Joung et al., 2020), which is related to three optical properties of chromophores, as well as 5 other datasets, i.e., MNSol (Marenich et al., 2020), Free Solv (Mobley & Guthrie, 2014), Comp Sol (Moine et al., 2017), Abraham (Grubbs et al., 2010), and Combi Solv (Vermeire & Green, 2021), which are related to the solvation free energy of solute.
Dataset Splits Yes For the molecular interaction prediction task, we evaluate the models under 5-fold cross validation scheme following the previous work (Pathak et al., 2020). The dataset is randomly split into 5 subsets and one of the subsets is used as the test set while the remaining subsets are used to train the model. A subset of the test set is selected as validation set for hyperparameter selection and early stopping.
Hardware Specification Yes We conduct all the experiments using a 24GB NVIDIA Ge Force RTX 3090.
Software Dependencies No The paper mentions using specific models like MPNN, GCN, GIN, and Set2Set, and the Adam optimizer, but does not provide specific version numbers for the underlying software libraries or programming languages used for implementation (e.g., Python version, PyTorch/TensorFlow version).
Experiment Setup Yes Hyperparameter details are described in Appendix E. [...] Table 7: Hyperparameter specifications (d: embedding dim, K: batch size, lr: learning rate, β: beta, τ: temperature). [...] We use the Adam optimizer for model optimization. For molecular interaction task and drug-drug interaction task, the learning rate was decreased on plateau by a factor of 10 1 with the patience of 20 epochs following previous work (Pathak et al., 2020).