|
<< Click to Display Table of Contents >> Navigation: Building blocks of GeNIe > Saving and loading models in GeNIe > Ergo file format |
Ergo file format was originally implemented by Noetic, Inc. in their implementation of a Bayesian network development environment, known as Ergo 1.0. It was quickly and in a somewhat ad-hoc manner embraced by various researchers in the Uncertainty in Artificial Intelligence (UAI) community because of its simplicity. Several important models developed in the first years of existence of the field of UAI were developed using Ergo format. Noetic, Inc., the developers and marketers of Ergo, have since changed their file format (we have implemented the format defined in Ergo version 1.02; we approached Noetic, Inc., at some point for a specification of the new format but have received no response) and they seem to no longer support their original format. This original, simple format has still survived in terms of useful Bayesian network models. The file extension, after Noetic, Inc., is *.erg.
The format supports only Chance nodes. Only node identifiers, state names, conditional probability tables, and locations of the node centers are saved. You will lose all other information. There is no description of the format available on-line but you should be able to figure it out by looking at some simple models. Here is the content of the Ergo file for the Venture BN example used throughout this document:
2
2 3
0
1 1
/* Probabilities */
2
0.2 0.8
6
0.4 0.4 0.2 0.1 0.3 0.6
/* Names */
Success Forecast
/* Labels */
Success Failure
Good Moderate Poor
/* Centers */
98 159
98 253
The first line in the file states how many nodes the model contains (in this case, 2). This is followed by the number of states of each of the node (2 and 3). The next lines state the parents of each of the nodes (the first node has 0 parents and the second node has 1 parent, node 1).
The Probabilities section lists the contents of the conditional probability tables (CPTs). The number that precedes each table is the number of parameters in each table.
Finally, the Names and Labels contain the node IDs and state IDs. Centers are coordinates of the centers of each of the nodes.