roywright/pu_learning

a problem in pu two step

Opened this issue · 3 comments

xuhongzuo commented
a problem in pu two step
xuhongzuo commented

Hi @roywright

range_P = [min(pred * (ys > 0)), max(pred * (ys > 0))]

I am wondering whether this line will directly assign the minimum value of positive data score as 0.

  • 👀1
BlingBlingss commented

Hi @roywright

range_P = [min(pred * (ys > 0)), max(pred * (ys > 0))]

I am wondering whether this line will directly assign the minimum value of positive data score as 0.

It is true,also

STEP 2

# Retrain on new labels and get new scores
rf2.fit(X, ys)

ys in STEP 2 will include -1(unlabeled)

Simon756015 commented

@xuhongzuo Hi, I also found that the minimum value will be directly set to 0. So does this code need to be changed?