In order to quantitatively evaluate the performance of the PCA-CMI and IPCA-CMI algorithms, we use simulated dataset which generated from Dream3 (Dialogue for Reverse Engineering Assessments and Methods) size of 10, 50 and 100 and real dataset (SOS DNA repair network with experiment data set in Escherichia coli).

In order to improve the PCA-CMI algorithm, we introduce an algorithm, IPCA-CMI, for learning the structure of the BN.

IPCA-CMI algorithm is written in Matlab. It gets simulated data and real data as inputs. This algorithm applied for learning the skeleton of BN.

How To Run The Program:

 

1.      One file is available for download (Dataset.rar): These file contains the data sets and true networks.

2.      One file is available for download (Java_Codes.rar): These file contains Java code for computing MIT scores. The MIT score belongs to the family of scores based on information theory. The MIT score is implemented within the Elvira system (a java package for Bayesian networks). The Elvira package can be downloaded from http://leo.ugr.es/elvira/. The MIT score is in the path\Java Codes\Elvira\bayelvira2\elvira. In this work we rewrite the MIT score program (Red.Pen) to improve the running time and memory occupied by the algorithm. Using the Red.Pen can reduce running time and memory.

3.      One file is available for download (Matlab_Codes.rar): These file contains Matlab codes for learning the skeleton of DAG.

4.      Run the program for each data set.