XBIF Format Specification
=========================

Introduction
------------

This document describes an XML format for BN specification.
Is is a variation of the original XMLBIF format proposed
by http://www.cs.cmu.edu/afs/cs/user/fgcozman/www/Research/InterchangeFormat/


Syntax
------

<xbif version="0.3a">
  <network>
    <name>'put here the instance name'</name>
    
    <!-- Variables -->
    
    <variable type="discrete">
      <name>'put here the variable name'</name>
      <values>'put here the domain size'</values>
      <observed>'put here the index of the observed
      		 value if the variable is part of the
      		 evidence set'</observed>
    </variable>
  
    .
    .
    .
    
    <!-- Probabilities -->
    
    <probability>
      <for>'put here the name of the child node'</for>
      <given>'put here the name of the 1st parent node'</given>
      <given>'put here the name of the 2nd parent node'</given>
      .
      .
      .
      <given>'put here the name of the ith parent node'</given>
      <table>
      	'put here the content of the probability table in
         form of a m x k matrix (columns are 'space/tab'
         separated), where m is the number of value combinations
         for the parents and k is the domain size of the child.
         The value combinations of the parents are considered in
         lexicographic order of the value indexes (see the Examples
         section below)'
      </table>
     </probability>
     
     .
     .
     .
     
  </network>
</xbif>


Conditions
----------

In addition it is assumed that the specification satisfies the 
following conditions:

1. The names of the variables are unique identifiers formed with
   characters from ['a-z', 'A-Z', '0-9', '_', '-'] and must begin
   with a letter.
   
2. The domain size is a positive integer from 1 to MAX_INT. The
   values are assumed indexed from 0 to domain size minus 1.

3. The observed value of an evidence variable is an integer from
   0 to k-1, where k is the domain size.
   
4. A probability table entry is a real number with a maximum of 8 
   digit resolution.

5. The XML tag names (e.g. <network>, etc.) can be either upper 
   or lower case. 

Examples
--------

Example: A simple network with 3 nodes: node_A, node_B and node_C
	 with domain sizes of 3, 2 and 3, respectively. The node 
	 'node_A' is evidence having its second value in the domain
	 as the observed value.

<xbif version="0.3a">
  <network>
    <name>Simple network</name>
    
    <!-- Variables -->
    <variable type="discrete">
      <name>node_A</name>
      <values>3</values>
      <observed>1</observed>
    </variable>
        
    <variable type="discrete">
      <name>node_B</name>
      <values>2</values>
    </variable>
      	      
    <variable type="discrete">
      <name>node_C</name>
      <values>3</values>
    </variable>

    <!-- Probabilities -->  
    <probability>
      <for>node_A</for>
      <table>
      	  0.2 0.3 0.5
      </table>
    </probability>
    
    <probability>
      <for>node_B</for>
      <table>
      	  0.3 0.7
      </table>
    </probability>

    <probability>
      <for>node_C</for>
      <given>node_A</given>
      <given>node_B</given>
      <table>
      	  0.2 0.3 0.5	// A=0 B=0
      	  0.3 0.4 0.3	// A=0 B=1
      	  0.1 0.8 0.1	// A=1 B=0
      	  0.2 0.4 0.4	// A=1 B=1
      	  0.4 0.5 0.1	// A=2 B=0
      	  0.3 0.3 0.4	// A=2 B=1
      </table>
    </probability>
  </network>
</xbif>