BIF-BNT converter

Use this Web service to convert a Bayes net from BIF format to BNT format. See below for more information on these data formats.

Upload the BIF file you wish to convert:

Alternatively, enter or paste BIF code below:
network Grass {} variable Cloudy { type discrete[2] {false true}; } variable Sprinkler { type discrete[2] {false true}; } variable Rain { type discrete[2] {false true}; } variable WetGrass { type discrete[2] {false true}; } probability (Cloudy) { table 0.5 0.5; } probability (Sprinkler | Cloudy) { table 0.5 0.9 0.5 0.1; } probability (Rain | Cloudy) { table 0.8 0.2 0.2 0.8; } probability (WetGrass | Rain Sprinkler) { table 1.0 0.1 0.1 0.01 0.0 0.9 0.9 0.99; }

(This may take a few seconds.)

(The default BIF code above is adapted from the BNT documentation, which in turn adapted the example from Russell and Norvig, "Artificial Intelligence: a Modern Approach", Prentice Hall, 1995, page 454.)

More on the data formats

By BIF format we mean Fabio Cozman's Interchange Format for Bayesian Networks (BNIF), used by the Bayesian Network Repository to distribute Bayes nets. BIF format is conceptually similar to, but distinct from, XMLBIF and XBN. For parsing convenience and robustness, our converter interprets the BIF format specification as follows:

All non-BIF material in the file, for example human-readable text, must be enclosed in comment markers.
The (unique) network block must appear at the very beginning of the file.
Following the network block, variable and probability blocks may appear interspersed. However, each probability block, relating a child variable to zero or more parent variables, must appear after the variable block(s) defining these variables.
Non-negative reals (i.e., probabilities) are allowed to not contain a decimal point.

By BNT format we mean MATLAB code that can be executed to create a Bayes net under Kevin Murphy's Bayes Net Toolbox (BNT). More specifically, this converter returns a MATLAB script that, when run with BNT in the MATLAB path, creates 4 variables in the current workspace:

bnet: The Bayes net represented by the input BIF file. The number of possible values for each (discrete) variable is stored in the row vector bnet.node_sizes.
adjacency: The adjacency matrix of the network, equal to bnet.dag.
node: A structure array whose field names are the names of each variable (wherein hyphens are replaced with underscores), and whose field values are the (topologically sorted and 1-based) indices of each variable.
value: A row cell array of cells, each of which is a column cell array of strings corresponding to a network node, listing the names of each possible value of the node.

Ken Shan <ken@digitas.harvard.edu>
December 28, 2000