syr-cn / SimSGT

[NeurIPS 2023] "Rethinking Tokenizer and Decoder in Masked Graph Modeling for Molecules"
27 stars 3 forks source link

About hyper-paramter tuning #5

Closed gihanpanapitiya closed 8 months ago

gihanpanapitiya commented 8 months ago

Did you do hyper-parameter tuning for each of the 10 random seeds? Or did you hyper-parameter tune using a particular seed and train the model using another 10 different random seeds to get the test set predictions?

syr-cn commented 8 months ago

Thanks for your interest.

For each particular hyper-parameter setting, we ran the experiments on the 10 random seeds and used the average performance on the 10 seeds to select the best hyper-parameter setting. We did not do hyper-parameter tuning for each seed separately.