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

Learning the tree from lifted relational random walks #39

Open asd36952 opened 4 years ago

asd36952 commented 4 years ago

Dear all,

After runing the lifted relation random walks, I get RWRPredicates.txt like following image

When I follow the guide of grounded relational random walks, the result is just inference with given weight.

If I run the with -l flag like,

java -cp boostsrl_v1.1.1.jar edu.wisc.cs.will.GroundRelationalRandomWalks.RunGroundRelationalRandomWalks -grw -mln -l -train "./data/NELL/train/" -target teamplayssport -trees 1 -model "./data/NELL/rwr_model/"

, than the code generates tree with new structure and weight.

I want to learn the weight for the tree from lifted relational random walks Is there any way to learn only weight without change the structure?

Thank you, -Seongwoo

navdeepkjohal commented 4 years ago

Hello Seongwoo,

Thank you for your interest in the random walks code. Currently, we do not support learning parameters on the random walks. However, we are working on the solution. Meanwhile, you could try to run alchemy-parameter learning module on the random walk structures (given our learnt structures are consistent with alchemy i/p), as an alternative

Best Navdeep