About the code on find_k_largest

[ll0ruc]

QRec/util/qmath.py

Line 138 in 2a13e3d

n_candidates.sort(key=lambda d: d[1], reverse=True)

the n_candidates is the top K index and values. the candidates[0:K] will append two times in the for, so I think for iid,score in enumerate(candidates), the iid and score should skip the 0~K-1?

[Coder-Yu]

It doesn't. I just checked the code and printed the list. It works as expected. Why do you think it will append two times?

[ll0ruc]

The for iid,score in enumerate(candidates means check the every value in candidates. While if the value is already in initial n_candidates and if the for, it will append one time again to the n_candidates. Just as the example find_k_largest(5, [1,2,3,4,5,6,7]). If the 0~K item in candidates is not the top K values, it will not affect the result.

[Coder-Yu]

QRec/util/qmath.py

Line 146 in 2a13e3d

if k_largest_scores[r] < score:

r=K-1 which means that for the first K items in candidates, the following codes conditioned on L146 would not be excuted. So, there would not be repeated items in the recommendation list.

However, it would be better if it iterates in the range [K:]. Thanks. I will consider to modify it.