I discovered and fixed a few bugs already in the code that approximates
non-exponential distributions with phase-type distributions, which makes me
doubt that this feature of Ymer 3 ever worked properly. The results from the
numeric computation of probabilities still seem suspicious, so it is likely
that the approximation code still is buggy.
Original issue reported on code.google.com by hlsyou...@gmail.com on 29 May 2012 at 12:59
Original issue reported on code.google.com by
hlsyou...@gmail.com
on 29 May 2012 at 12:59