Testing the Module (BDeu)

This module implements Equation 35 of Heckerman's article in Theory & Concepts

Testing the Module (BDeu)

Postby M Charles » Wed Aug 06, 2014 5:55 pm

I am still verifying the correctness of this module. Anyway, here is one quick test.

The test uses 1000 records for a 3-node network of binary states. The data was generated from a network with no arcs and different probabilities (Node1 had 0.75:0.25, Node2 had 0.5:0.5, Node3 had 0.4:0.6).

The following image contains the scoring results of 25 different 3-node networks with the full double-floating point precision.

test.jpg
test.jpg (247.72 KiB) Viewed 1450 times
M Charles
 
Posts: 23
Joined: Sun Jun 22, 2014 5:00 pm

Re: Testing the Module (BDeu)

Postby M Charles » Wed Aug 06, 2014 7:28 pm

Here is another test with only 3 records:
Node1 Node2 Node3
1 1 0
0 0 0
1 0 1


test.jpg
test.jpg (248.45 KiB) Viewed 1447 times
M Charles
 
Posts: 23
Joined: Sun Jun 22, 2014 5:00 pm

Re: Testing the Module (BDeu)

Postby cwyoo » Wed Aug 06, 2014 9:09 pm

Michael Cho wrote:I am still verifying the correctness of this module. Anyway, here is one quick test.

The test uses 1000 records for a 3-node network of binary states. The data was generated from a network with no arcs and different probabilities (Node1 had 0.75:0.25, Node2 had 0.5:0.5, Node3 had 0.4:0.6).

The following image contains the scoring results of 25 different 3-node networks with the full double-floating point precision.

test.jpg


All equivalence classes seems correct,. Nice work. Did you also test on a larger (in terms of number of variables) dataset with a random structure (number of parents limited to 5)? You might use the hypertension genes data.
cwyoo
Site Admin
 
Posts: 379
Joined: Sun Jun 22, 2014 2:38 pm

Re: Testing the Module (BDeu)

Postby cwyoo » Wed Aug 06, 2014 9:20 pm

Michael Cho wrote:Here is another test with only 3 records:
Node1 Node2 Node3
1 1 0
0 0 0
1 0 1


test.jpg


Try using the following data to verify the score:
Node1 Node2 Node3
1 1 0
0 0 1
1 0 1
cwyoo
Site Admin
 
Posts: 379
Joined: Sun Jun 22, 2014 2:38 pm

Re: Testing the Module (BDeu)

Postby qxin001 » Mon Jun 12, 2017 4:36 pm

After compiling chronic disease model on path two, the result like below:

qxin001@path-two:~/Desktop/Test$ cd chronic-disease-model.git/
qxin001@path-two:~/Desktop/Test/chronic-disease-model.git$ make clean
rm -rf build bin/model lib/libbnl.a
qxin001@path-two:~/Desktop/Test/chronic-disease-model.git$ make
g++ -isystem thirdparty/linux_x64/include -Iinclude -O0 -DDEBUG -g -Wall -Wextra -Werror -pedantic -c -o build/bnl/refcounted_dataset.o src/bnl/refcounted_dataset.cpp
In file included from include/bnl/random/algorithm/random_shuffle.hpp:10:0,
from src/bnl/refcounted_dataset.cpp:2:
include/bnl/random/randgen.hpp:16:45: fatal error: boost/random/mersenne_twister.hpp: No such file or directory
#include <boost/random/mersenne_twister.hpp>
^
compilation terminated.
make: *** [build/bnl/refcounted_dataset.o] Error 1
qxin001
 
Posts: 2
Joined: Wed May 10, 2017 12:01 pm


Return to BDe Scoring Module

Who is online

Users browsing this forum: No registered users and 2 guests