Closed mattbarr-g closed 10 years ago
Just a clarification, the error only occurs when input is provided to the Discrete making up the number of trials, not the when input is provided to the Discrete making up the success count.
Internally this is bug https://bugzilla.cdcdesign.analog.com/show_bug.cgi?id=83
Fixed
I receive an 'Energy value is NaN' about half of the time when using Gibbs on a simple factor graph using a binomial factor in the Matlab interface. The other half of the time, it samples fine and I get beliefs roughly equivalent to the BP solve (which works fine).
Simplified Matlab code directly below and error included afterward.
Error message: