Questions Regarding Hyperparameter Settings and Score Thresholds for Figure 8 Reproduction #1
Description
Hi,
First of all, thank you for your great work on this project. I have a couple of questions regarding the reproduction of Figure 8 (Learning curves of affinity scores where the red dashed line represents the threshold):
When reproducing the grouped results in Figure 8, should the hyperparameters be set according to the default parameters in the add_args function in utils.py or according to the parameters provided in molgroup/scripts/dataset_grouping.sh?
Additionally, I have a question about line 476 in molgroup/app/train_molgroup.py. The comment mentions "# get the index of gate with the score not higher than 0.4", which seems to suggest that only tasks with gate_scores not higher than 0.4 are retained. However, the paper mentions that in each round, auxiliary datasets with affinity scores lower than 0.6 are removed. Is this difference due to a distinction between the definitions of gate_scores and affinity scores?
Thank you for your assistance!
Best regards