question about random data split

[xinqiao123]

Hello zilong,

Thanks for your excellent research work.

I have a question about random training data split. In your paper, you mentioned every model was run 5 times, from your code, each run using a same percent but different samples, with shuffle indices, that make sense to me.

What I feel confused is when we compare different models, are you also using the shuffle indices? For example, if we just run once for all models, we use the same data or different data to compare the preformance of different models?

Thanks for your time,
Xin