a7b23/H-Divergence

Regarding fixed seed for each trial

Closed this issue · 1 comments

Piyushi-0 commented

For sampling s1, s2 in the blobs experiments, the same seed is being used np.random.seed(seed=11 * k + 10 + n) in each trial. Ref.

H-Divergence/blob_gmm.py

Line 105 in 19030d3

np.random.seed(seed=11 * k + 10 + n)
and

H-Divergence/blob_kde.py

Line 137 in 19030d3

np.random.seed(seed=11 * k + 10 + n)
It doesn't depend on the variable kk denoting the trial index. Could you please share the reason for this?

KellyYutongHe commented

We followed the experiment setup from Liu, et. al 2020 (https://github.com/fengliu90/DK-for-TST/blob/master/Baselines_Blob.py) and this is how they set the random seed.