BayesFusion Support Forum

Hi, The "prior link probability" parameter influences the likelihood of the Bayesian Search algorithm producing arcs, as higher values favor arcs. So, it is not really surprising that when you increased its value, you got a denser network. My answer to the question whether you can modify t...

Dear Marco, I doubt the automatic discretization of continuous variables will be of much help to you. What it does is turning a continuous variable into a discrete variable (producing a CPT for it based on the definitions of the parents and the node in question). Since you don't have a parametric de...

Hi, The EM algorithm is general enough to handle all possible cases in theory, some are easier and straightforward, other complex. It is just that we have not yet implemented anything outside the discrete case. We will do it at some point but in the next few months due to other priorities. I guess y...

Hi, Let me try to help, although I realize that I may not know enough to give useful suggestions. I know that you want to use discrete variables but have you considered continuous variables with Gaussian distributions and linear relationships? Because your scores are averages and are continuous, the...

No, I'm afraid there is nothing like that in GeNIe. The idea is interesting, though!

Yes, you could view it this way.
Cheers,

Marek

Hi Heyifan, What you wrote is essentially correct, although the mechanics of this is somewhat more sophisticated than reducing the existing parameters to 100 records. If we did that, there would be a chance of never having records for some combination of values of parent nodes. The way it is impleme...

Hi Sverre, Zeros are not dangerous because of the possibility of division by zero but rather multiplication. In Bayes theorem, the posterior is derived by multiplying the prior by an expression that represents the strength of evidence. If the prior is zero, then the posterior HAS TO be zero. For exa...

No missing values is just a special case for the EM algorithm, so there is no need for a special separate algorithm for data with no missing values. I hope this helps,

Marek

Hi Lotte, Glad to be of help. I don't have any reference handy but you can easily find references to cross entropy using Google search. I tried "cross-entropy" and "cross-entropy" application with many interesting results. "cross-entropy" "SMILE" BayesFusion w...

Hi Sverre, I don't believe I could answer your question general. My feeling is that the network with individual nodes will be better but to make an informed judgment I should see and examine the two networks. You are mentioning ease of getting the numbers. This is a very important factors, as speed ...

I will be happy to look at this problem. Would you be willing to share the piece of our model that produces this behavior?
Cheers,

Marek

There are plenty of references, too many to list. A Google search on "imbalanced dataset handling" will do the job -- you will find review articles that will recommend you some methods and offer references.
Good luck!

Marek

GeNIe does not offer any special treatment to imbalanced data at the moment. We are planning to return to this issue quite likely in mid-late 2022. At the moment, you will have to do the pre-processing (balancing, boosting, or whatever you choose) yourself before you read your data into GeNIe. I hop...

Hi HeyFan, Uniform distributions are as good as (or as bad as) a randomly chosen starting point and you should not view them as the best! To the contrary, it often leads to worse convergence. If you are worried about different convergence points, I suggest that you try a few randomizations and compa...