Please sign in to comment.
start at sample_proposal. v is wrong there
- Loading branch information...
Showing with 83 additions and 22 deletions.
|@@ -1,3 +1,13 @@|
|+even a little bit of numerical error results in exp(E) to go zero (when E<0) in gauss evaluate.|
|+The problem is the input to gauss evaluate, v. It's set in sample_proposal. line 172. v is wrong...|
|+Do a check:|
|+is likelhood_given_xv returning a 0 with the right inputs? Then, the function has a problem. If the inputs are wrong, then sample_particles might be wrong. Check this (add a break in likelihood_given_xv.|
|+w is zero in resample particles. w is set in sample_proposal. The likelihood given xv returns a 0 everytime, causing w to be 0. This results in nan's, which should throw a division by 0 error|
|fix the multivariate gauss's use of Random() (from eigen). It doesn't return diff. random numbers at each turn|