if snorkel support soft-labelling function

[futianfan]

Hi,

I notice that all these labelling function in snorkel can only produce discrete label, but if i have a binary classification problem where labels are soft, i.e., 0.8 mean it is positive w.p. 0.8 and negative w.p. 0.2. can snorkel support this kind of soft-labels?

Thanks very much in advance!

[vincentschen]

Good question—this is on our roadmap related to regression LFs.

Closing and tracking on #1072

[prpankajsingh]

@futianfan You can refer the work done by IIT Bombay Link
This work extends the snorkel framework and output a continuous score (in-stead of a hard label) that noisily correlates with labels