W3C home > Mailing lists > Public > www-archive@w3.org > November 2006

Re: additional bayesian test case [Was: 2 e:bayesian test cases]

From: <jos.deroo@agfa.com>
Date: Thu, 30 Nov 2006 00:43:39 +0100
To: www-archive@w3.org
Message-ID: <OFEC8982B9.15053D7B-ONC1257235.007FB4A1-C1257235.00824AFA@agfa.com>
.. and using bgot/3 gives another speedup of 350 times which is now
45000 times faster in 2 days or from 377,561,246 to 8,289 inferences
for a test case using 70 Bayesian Net nodes and I think that
makes N3 based BN inferencing quite feasable, esp. because the
logical case is a subcase using logical variables with probabilities
0 and 1 only @@ http://eulersharp.sourceforge.net/2004/04test/gate.n3

> .. there was an approximation in that version which is now fixed (*),
> with inspiration from http://www.cis.cau.edu/475/ibratko/fig15_11.pl
> and speed for 30 node graph is boosted with factor 130 using memo/0
< in http://eulersharp.sourceforge.net/2006/02swap/euler.yap

>> also found Yoshio's
>> http://www.w3.org/2004/09/13-Yoshio/PositionPaper.html
>> and the case in there using
>> http://eulersharp.sourceforge.net/2003/03swap/log-rules speak is
>> GIVEN http://eulersharp.sourceforge.net/2004/04test/metastatic.n3
>> QUERY http://eulersharp.sourceforge.net/2004/04test/metastaticF.n3
>> PROOF http://eulersharp.sourceforge.net/2004/04test/metastaticE.n3

--
Jos De Roo, AGFA http://www.agfa.com/w3c/jdroo/
Received on Wednesday, 29 November 2006 23:43:54 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 7 November 2012 14:18:00 GMT