danfu09

4 ms·

1. 1.▲
   by danfu09 7y ago

   We can only do classification (for now), and we need users to be able to provide enough labeling functions that are correlated well with the true labels. Text and some video tasks are examples of where it can work well, since it's poss
2. 2.▲
   Towards Interactive Weak Supervision with FlyingSquid(hazyresearch.stanford.edu)

   75 points by danfu09 7y ago|6 comments
3. 3.▲
   Software 2.0 and Data Programming: Lessons Learned, and What’s Next(hazyresearch.stanford.edu)

   6 points by danfu09 7y ago|0 comments
4. 4.▲
   Why Train What You Can Code? Rekall: A Compositional Approach to Video Analysis(dawn.cs.stanford.edu)

   7 points by danfu09 7y ago|0 comments