Hi,
I have ecountered the following bug:
I load a network from an XBN file, the loading goes well but when I do inference I got the error reported below.

You can find the XBN file alarmc.xbn at
http://www.ing.unife.it/docenti/FabrizioRiguzzi/alarmc.xbn

Any idea what happened?

Thanks,
Fabrizio

Error:
-------------
Traceback (most recent call last):
File "./testalarmc.py", line 9, in
ie=JoinTree(G)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 426, in __in
it__
self.GlobalPropagation()
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 514, in Glob
alPropagation
start.CollectEvidence()
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 187, in Coll
ectEvidence
if not v.marked: v.CollectEvidence(self)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 187, in Coll
ectEvidence
if not v.marked: v.CollectEvidence(self)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 187, in Coll
ectEvidence
if not v.marked: v.CollectEvidence(self)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 187, in Coll
ectEvidence
if not v.marked: v.CollectEvidence(self)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 187, in Coll
ectEvidence
if not v.marked: v.CollectEvidence(self)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 189, in Coll
ectEvidence
if not X == None: self.MessagePass(X)
File "C:\Python25\Lib\site-packages\OpenBayes\inference.py", line 168, in Mess
agePass
newphiR = self.potential + e.potential # phiR = sum(X/R)phiX
File "C:\Python25\Lib\site-packages\OpenBayes\potentials.py", line 154, in __a
dd__
return self.Marginalise(var)
File "C:\Python25\Lib\site-packages\OpenBayes\potentials.py", line 117, in Mar
ginalise
return DiscretePotential(newnames, temp.shape, temp)
AttributeError: 'float' object has no attribute 'shape'
-------------
File testalarmc.py:
------------
#! /usr/bin/env python
from OpenBayes import LoadXBN,JoinTree,MCMCEngine
from time import clock

l=LoadXBN("alarmvar.xbn")
G=l.Load()

t=clock()
ie=JoinTree(G)
print ie.Marginalise('alarm_t_')
t=clock()
print t

Hi Fabrizio,

I don't think OpenBayes has been tested extensively under Python2.5 yet.

I ran your test script against your XBN file and I got the successful output:

INFO:root:Creating JunctionTree engine for bndefault
INFO:root:Constructing Optimal Tree
INFO:root:Moralising Tree
INFO:root:Triangulating Tree and extracting Clusters
INFO:root:Connecting Clusters Optimally
INFO:root:Initialising Potentials for clusters and SepSets
INFO:root:Global Propagation, starting at :earthq_f_c_5_
INFO:root: Collect Evidence
INFO:root: Distribute Evidence
Multinomial Distribution for node : alarm_t_
Conditional Probability Table (CPT) :
array([ 0.86200321, 0.13799682], type=Float32)
0.49

Hope this helps!

Sébastien

Hi,
thanks for you quick reply.
I tried the same script with python 2.4 on a windows machine and on a linux machine and I get the same error.
Are you using a development version of OpenBayes?
I downloaded my version from the dist direcotory.

Thanks,
Fabrizio

Hi,

I am not sure which version I am running I downloaded whichever one was last released in December.

You might want to try an older revision:
http://svn.berlios.de/wsvn/pybayes/branches/Public/dist/OpenBayes-0.1.0.tar.gz?op=log&rev=0&sc=0&isdir=0

I hope that Kosta is considering using a python-egg to distribute OpenBayes at some point, so that it becomes much easier to upgrade the package and manage different releases.

Thanks,

Sébastien

Hi,
sorry I made a mistake, the file that gives me the error is not alarmc.xbn but alarmvar.xbn,
here it is
http://www.ing.unife.it/docenti/FabrizioRiguzzi/alarmvar.xbn
Here is the python script that gives me the error
http://www.ing.unife.it/docenti/FabrizioRiguzzi/alarmvar.py

alarmc.xbn works also for me

Thanks,
Fabrizio