• aolibPE
• Demo

aolibPE step-by-step instructions

In the following we provide step-by-step instructions on how to use the aolibPE algorithm on an example problem instance. This example is based on Windows, but the steps for Linux are almost the same.

Downloading and running the algorithm

1. Download the archive with the executable from the algorithm's page. Since we assume a Windows operating system here, we need this file.

2. Unpack the archive contents (the binary and a sample parameter file) into a directory of your choice, say C:\aolib\.

3. Obtain an Ergo file of a P(e) problem instance, for instance from our local Repository. For this example we pick BN_11 from the UAI'06 competition. Don't forget to download the respective evidence file, too.

4. The problem instance and evidence file come in gzip'ed form, so unpack them to the same directory as the algorithm executable; in our case C:\aolib\.¹

5. Open a command line (In Windows: Choose "Run" in the start menu and enter cmd) and navigate to the directory where we put the files by entering cd C:\aolib\.

6. Run the algorithm by specifying the problem file, the evidence file, the parameter file and the output file, as described on the algorithm's page. We'll use the parameter file that comes with the program file. If we wish to have the result output into the file BN_11.out, the command line call in our case will be (all on one line):

   aolibPE
       BN_11.erg
       BN_11.erg.evid
       paramsPE.txt
       BN_11.out

Algorithm output

The algorithm will run and, depending on the speed of your computer, the result should be returned shortly afterwards. The output should look somewhat like this:

[elim = 10]
[iBound = 22]
[be = 1]
[usePruning = 0]iBound = 22
usePruning = 0
SWITCH TO BE = true
file = BN_11.simple
 Starting to read file ... 
 Done reading file. 

 initial number of variables   : 105
 number of non-barren variables: 55
 scaling factor = 1.01423e+304 Induced width: -1

 BEGIN AO SEARCH  -----------------------------------------
   time limit: -1
   silent mode: true
   DETAILS
    -int: time -0.79646, nodes 0       done: 0 %
   out of time: false
   component probability of evidence: 3.06314e+303
   CPU time (search):          4.70867e-05
   number of expansions:        0
   number of AND expansions:    0
   number of OR expansions:     0
 END AO SEARCH    -----------------------------------------

 scaling factor = 426327 Induced width: 15

 BEGIN AO SEARCH  -----------------------------------------
   time limit: -1
   silent mode: true
   DETAILS
    -int: time -0.813329, nodes 0       done: 0 %
   out of time: false
   component probability of evidence: 1.69837e+294
   CPU time (search):          0.522059
   number of expansions:        0
   number of AND expansions:    0
   number of OR expansions:     0
 END AO SEARCH    -----------------------------------------

Log probability of evidence: -39.3718
TIME: 
   - Read file time:       0.0100231
   - Preprocessing time:   0.0222144
   - Solving time:         0.52244
                         --------------------------------
   - TOTAL                 0.554951

In addition to this diagnostic output, the log probability of evidence has been written to the file BN_11.out. aolibPE also outputs (or appends) a summary of its computation to the files shortResults.txt and longResults.txt, which you should feel free to look at.