search

starling-lab / BoostSRL

BoostSRL: "Boosting for Statistical Relational Learning." A gradient-boosting based approach for learning different types of SRL models.
https://starling.utdallas.edu
GNU General Public License v3.0
32 stars 21 forks source link

Initial potential #27

Open rodrigoazs opened 5 years ago

rodrigoazs commented 5 years ago

Hello,

It was mentioned, in paper Gradient-based boosting for statistical relational learning: The relational dependency network case, that the initial potential had to be set to -1.8 in order to capture the uniform distribution. Where does this value -1.8 come from?

Thank you.

hayesall commented 5 years ago

This thread might be helpful: https://github.com/starling-lab/rfgb.py/issues/7