BDEu learning- weigths param

frito · Post by **frito** » Wed Aug 27, 2008 10:55 pm

What does the weights parameter mean, it definitely does generate different structures?

mark · Post by **mark** » Fri Aug 29, 2008 4:01 am

It's the equivalent sample size associated with the Dirichlet distribution which determines the prior over the parameters. Higher values will enforce more uniformly distributed CPTs.

Ken · Post by **Ken** » Wed Oct 15, 2008 1:22 am

What do you recommend to be the basis for weighting the Dirichlet prior information when we update with data? In other words, how should we choose the alpha parameters for the prior?

mark · Post by **mark** » Wed Oct 15, 2008 2:37 am

I would recommend the default size used in GeNIe, namely 1. You should probably also play a bit with the parameter and see what happens.