Bug? single labeling function ignored

[orenbookarang]

Issue description

I have a single LF. It seems to operate correctly, yet the labeling ignores it.

Code example/repro steps

Was Filmed - With Snorkel - 23 Aug 2019(1).pdf

Expected behavior

Behaves somewhat like the single LF. Not always -1.

Screenshots

See attached PDF (notebook)

System info

• How you installed Snorkel (conda, pip, source):
• Build command you used (if compiling from source):
• OS: Darwin Black-Box.local 17.7.0 Darwin Kernel Version 17.7.0: Wed Apr 24 21:17:24 PDT 2019; root:xnu-4570.71.45~1/RELEASE_X86_64 x86_64
• Python version: Python 3.6.4
• Snorkel version: snorkel 0.9.0
• Versions of any other relevant libraries:

Additional context

Add any other context about the problem here.

[orenbookarang]

By now I have discovered predict_proba, but also there it returns always [0.5, 0.5].

The next progress I had was that I've found that if my LF(s) return 0 instead of -1 in the "unknown" case, I do get interesting and relevant probabilities (not just 0.5).

Still will be happy for clarification/help or potential improvement as to make it better self explanatory.

[paroma]

The LabelModel requires at least 3 labeling functions (LFs) as input. This ensures the model has enough overlaps and conflicts to learn the parameters associated with LFs. With a single labeling function, the label model cannot learn these parameters and the outputs are therefor invalid. More details on the how the LabelModel functions here

Tracking the issue with multiple LFs still causing equal probabilities given the output is either an abstain or class 1 for a datapoint in #1422. Closing this for now, feel free to reopen with any questions!