Open mglowacki100 opened 1 year ago
Hi @mglowacki100,
Thanks for proposal. Do you get score iprovement when applied min, max golden features?
@pplonski I didn't test it, yet. I think the most sensible way to check it is to extract this to separate package and run across e.g. openml datasets and observe what types of golden features are at the top for most datasets. What do you think about it?
Yes, that would be proper approach. It is quite big project.
That looks good, while we are at it whats about sqrt and pow? .
I have another question as i get it this line. new_cols_cnt = np.min([100, np.max([10, int(0.1 X.shape[1])])]) the 0.1 saying that we get max 0.1 original column size as new futures, would be nice if we are able to change this .
I think it would be "low hanging fruit" to add min and max golden features. As far I see, it requires changes in three places:
Btw. maybe 250000 should be configurable? What do you think about it, below is implementation proposal: