search

gtjemwa / augustus

Automatically exported from code.google.com/p/augustus
0 stars 0 forks source link

GLR implementation possibly incorrect #32

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago 
The GLR implementation in PMMLBaseline.py does not seem to agree with the 
derivation in "Detection of Abrupt Changes: Theory and Application" at the 
bottom of page 53. The summed quantity on line 532 of PMMLBaseline.py should be 
a square of sums rather than a sum of squares.

Original issue reported on code.google.com by knoa...@gmail.com on 27 Jul 2010 at 2:00

GoogleCodeExporter commented 8 years ago 
I think you are right.  I'll take a look at the impact this has on the GLR code.

There is a release scheduled for August, and I'd rather not push any bug fixes 
before then so that everything can be tested together.  I will try to get this 
and the last few reported bugs included in the release.

Original comment by collinbe...@gmail.com on 29 Jul 2010 at 2:32

GoogleCodeExporter commented 8 years ago 
Ok -- it might also be good to modularize the code so that GLR tests can be 
added for non-Gaussian distributions.

Original comment by knoa...@gmail.com on 29 Jul 2010 at 9:31

GoogleCodeExporter commented 8 years ago 
After spending some more time reviewing the code, there was something wrong, 
but it wasn't that the sum of squares was being computed.  We have a fix and an 
explanation, I'll try to write it up and put it in the documentation on this 
site in the next few days.

Original comment by collinbe...@gmail.com on 29 Aug 2010 at 3:37

GoogleCodeExporter commented 8 years ago 
Addressed with augustus/pmmllib/pmmlBaseline.py revision 263.
I hope to get something on the wiki about this soon.

Original comment by collinbe...@gmail.com on 30 Aug 2010 at 7:16