search

DearCaat / MHIM-MIL

[ICCV 2023 Oral] Multiple Instance Learning Framework with Masked Hard Instance Mining for Whole Slide Image Classification
49 stars 3 forks source link

seed setting #10

Closed lingxitong closed 4 months ago

lingxitong commented 4 months ago

in your paper ,I see you rerun 3 times and report the std and mean value,that is good use to avoid random ans,But I have a quest,should the dataset split by seed ,such as you use seed 2021 2022 2023 to run three times ,the dataset split in the three times is different or same?can you tell me which is good? thanks

DearCaat commented 4 months ago

The seed setting in the code changes all packages and python3's random seed. Therefore, the data split is different on all 3 times. I can't say for sure which one is optimal, but I personally think this allows me to evaluate the performance of the model with different data splits.

lingxitong commented 4 months ago

thanks