What's the sample strategy of BPR?

[yangxin42]

I trained a dataset(ml-100k) using the BPR algorithm for serval times, and every time the evaluation scores are the same. So I figure what sample strategy the BPR algorithm takes? Thanks for your assistance!