Issue 751 in beast-mcmc: Full evaluation error

[beast...@googlecode.com]

Status: Accepted
Owner: mdhall272
Labels: Type-Defect Priority-Medium

New issue 751 by mdhall272: Full evaluation error
https://code.google.com/p/beast-mcmc/issues/detail?id=751

Run the attached with seed 1416263702625. You should get a full evaluation
error at state 1193:

State 1193: State was not correctly restored after reject step.
Likelihood before: -139383.49299576558 Likelihood after: -139383.4584840466
Operator: dr.evomodel.operators.ExchangeOperator@2594e776 Wide
Exchange(treeModel)

Details
Before:
CompoundLikelihood(compoundModel)=(
LogNormal(kappa)=-2.2567,
Gamma(skygrid.precision)=-10.6715,
GMRFMultilocusSkyrideLikelihood(gmrfSkyrideLikelihood[skygrid])=-2208.0189
Total = -2220.947049730265
),
CompoundLikelihood(compoundModel)=(
TreeLikelihood(treeLikelihood)=-137162.5459
Total = -137162.54594603533
)
Total = -139383.49299576558
After:
CompoundLikelihood(compoundModel)=(
LogNormal(kappa)=-2.2567,
Gamma(skygrid.precision)=-10.6715,
GMRFMultilocusSkyrideLikelihood(gmrfSkyrideLikelihood[skygrid])=-2208.0189
Total = -2220.947049730265
),
CompoundLikelihood(compoundModel)=(
TreeLikelihood(treeLikelihood)=-137162.5114
Total = -137162.51143431634
)
Total = -139383.4584840466


This happens whether or not Beagle is enabled (although the "before"
likelihood is -Inf rather than -137162.54594603533 if it is on). It doesn't
happen with beagle_scaling always though.

Attachments:
unif_unif_counts_425_5_skygrid.xml 343 KB

--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings

[beast...@googlecode.com]

Comment #1 on issue 751 by ramb...@gmail.com: Full evaluation error
https://code.google.com/p/beast-mcmc/issues/detail?id=751

Looks like the dynamic scaling is loosing some precision. The difference is
at the 2nd decimal place in the treelikelihood. Try using -beagle_rescale
100 . This forces a full rescaling every 100 steps (the default is 1000).

[beast...@googlecode.com]

Comment #2 on issue 751 by mdhall272: Full evaluation error
https://code.google.com/p/beast-mcmc/issues/detail?id=751

No joy, unfortunately - the error occurs all the way down to
-beagle_rescale 1.

[beast...@googlecode.com]

Updates:
Labels: beast v1.8.1 Milestone-Release1.8.2

Comment #3 on issue 751 by ramb...@gmail.com: Full evaluation error
https://code.google.com/p/beast-mcmc/issues/detail?id=751

(No comment was entered for this change.)

[beast...@googlecode.com]

Comment #4 on issue 751 by ramb...@gmail.com: Full evaluation error
https://code.google.com/p/beast-mcmc/issues/detail?id=751

'-beagle_scaling always' fixes this issue confirming it is a rescaling
issue. When using default scaling, the error occurs after an underflow
gives a -Inf likelihood and then reevaluation with a rescale gives a
non-infinite likelihood.

[beast...@googlecode.com]

Comment #5 on issue 751 by ramb...@gmail.com: Full evaluation error
https://code.google.com/p/beast-mcmc/issues/detail?id=751

I understand why this is happening: An underflow is triggering a
recomputation of the scale factors but the likelihood is not being
recomputed correctly with the new scale factors leaving a -Inf. When the
full evaluation test happens it recomputes the full likelihood correctly
with the new scale factors. Not sure what the fix is yet.