Messages

1

Java / UnsatisfiedLinkError

« on: December 12, 2013, 17:47:28  »

Dear All,

I'm getting the following error when trying to run a Java application based on HUGIN (MacOS 10.8.5)

       UnsatisfiedLinkError: no hapi70 in java.library.path

The java.library.path is: ./java/build/classes:./java/lib/*
libhapi70.jnilib is present in the /java/lib/

Any suggestions?

Many thanks in advance,
ally789

2

General Discussion / Re: Independence of variables

« on: March 01, 2011, 15:11:10  »

Thank you very much for the quick answer!

I still worrying a bit about the quality of the resulting network. Let's assume the following:
I have three nodes A, B and C. I define A and B to be independent of each other. Structure learning reveals a link between A and C and another link between B and C.

What are the consequences for my network if A and B are actually *not* independent of each other (given the data set)? Do I violate any rules of BN usage this way?

Many, many thanks...

3

General Discussion / Independence of variables

« on: February 28, 2011, 10:31:09  »

Consider the following situation:

I have a set of data (discrete variables ) and apply NPC and EM learning. Before learning I define some constraints in terms of "no link constraint" to reduce the search space and to be able to achieve useful results given only relatively sparse data and quite some nodes...

What happens if I assume "no link" between two variables although they are actually not independent from each other given the data situation? Is this "illegal"?

Thank you very much for your help in advance!

Best,
ally789

4

C++ / Exception: "inconsistent evidence or underlow"

« on: February 06, 2010, 15:40:18  »

I'm getting the following exception when propagating evidence through my network (via C++ API):

     "Propagation of inconsistent evidence has been attempted, or underlow has occured"

Is there anybody who has idea what the reason for this might be?? When doing the same in the GUI everything is fine...

Thanks in advance,
ally789

5

General Discussion / Independence Test

« on: January 16, 2010, 15:28:32  »

Can anybody tell me which statistical tests for conditional independence are performed when running the PC/NPC algorithm?

In a post from 2007 I found the information that G2 is used. Now I have compared PC learning in Hugin with PC learning using the BNT implementation with G2:  And the resulting networks differ...

Many thanks in advance!

6

C++ / Re: parseCases fails

« on: July 16, 2008, 16:58:25  »

Hmm, that's really strange...
Everything else works fine for me, two. I can build a network "manually", I can propagate evidence etc. Only calling parseCases causes the crash.

7

C++ / parseCases fails

« on: July 16, 2008, 11:22:36  »

Hi,

I'm using the following code to enter a set of cases into a domain:
   
   const string netFileName = fileName + ".net";
   Domain myDomain (netFileName, NULL);
   myDomain.compile ();

   string dataFileName = fileName + ".dat";
   MyParseListener pl;
   myDomain.parseCases (dataFileName, &pl);

I'm using Asia.net and the following .dat-file (as in the Hugin Manual):

   # A S D X
   1 "yes" "no" "no" "no"
   1 "no" "yes" "yes" "no"
   1 "no" "yes" "yes" "yes"
   1 "no" "no" "yes" "yes"
   2 "yes" "yes" * "no"
   1 "yes" "no" "no" *
   1 "yes" "yes" "yes" "yes"
   1 "no" "no" "no" *

What happens is, that the application always crashes when parseCases is called... Maybe there is something wrong with the .data-file? Any ideas are welcome!

Thanks,
ally789