Gradient Ascent for Bayes Nets

Let 183#183 denote one entry in the conditional probability table for variable 181#181 in the network

184#184

e.g., if 185#185, then 186#186 might be 187#187

Perform gradient ascent by repeatedly

1. update all 183#183 using training data 5#5
   
   188#188
   
   

2. then, renormalize the 183#183 to assure
   • 189#189
   • 190#190