About the ratio of training samples that have Xt! = Xs

[kyugorithm]

Hi. Thank you for sharing a great code.
I'm just posting because there's a very small problem.
The author of the paper specifies that the probability of not being the same is 80%, and the probability of being the same is 20%. However, this code specifies the same probability as 80%. I think this part should be revised.

AS-IS

TO-BE

Thank you