How small probabilities are supported and what is the number of significant d...

[Anders L Madsen]

How small probabilities are supported and what is the number of significant digits?

The Hugin Graphical User Interface uses a double precision inference engine. That is, it uses "IEEE 754" double-precision floating point computations. This allows computation of numbers as small as about 1E-300.
IEEE 754 double-precision means 53 significant bits. This is equivalent to something like 15-16 decimal digits. Of course, the last few bits cannot be trusted because of possible rounding errors. However, the computations used for computing probabilities in networks with discrete nodes are very stable, so the first 13-14 decimal digits can be trusted.