Testing Larger Datasets

[M Charles]

Using the dataset inside the "Tree Structure" forum section (PAH-Fnal.xlsx - 146 records, 166 variables) on my 2.4GHz Intel Core 2 MacBook Pro, it took:

• Dataset construction: 0.00445104 sec
• Random network (max 5 parents): 0.0167201 sec
• Scoring and drawing DOT file: 32.9561 sec
• Writing GeNIe file: 0.16809 sec

The network's log score (w/ BDeu metric) was -29985.5286853225

-----

If I ignore the file writing stage (only compute score) and run on the server, I get:

• Dataset construction: 0.00357199 sec
• Random network (max 5 parents): 0.00900912 sec
• Scoring the network: 0.160258 sec

In this trial the (different) network's log score (w/ BDeu metric) was -30463.9