|
Hi I am looking for a simple framework that can learn structure from structure and CPT'. I am looking for a simple framework that can learn structure from structure and CPT' from data and should be scalable. Can someone chime in on any experiences they may have had with some implementations. I am using PEBL with python but it does not learn CPT's apparently. I wanted something in python but for the lack of frameworks in this field, a framework in any language that I can make python bindings for is also acceptable here... Also: 1) how can I prevent overfitting in the learning stage? 2) If the CPT's change with time somehow how can I incorporate that into the system? 3) An online algorithm that can update teh n/w would be nice. I am looking at weka which has a bayesian classifier. It was suggested that I just chose any arbitrary feature as the predictor Y to trick weka into learning the CPT's, does that seem like a resonable way to use weka? |
