Smilearn - tutorial 6 - why do we need A,B,C?

The engine.
Post Reply
timw
Posts: 7
Joined: Fri Apr 25, 2014 10:42 pm

Smilearn - tutorial 6 - why do we need A,B,C?

Post by timw »

In tutorial 6: Learning a Dynamic Network with EM we use a data set as follows:

A A_0 A_1 B B_0 B_1 C C_0 C_1

I understand that i_0 and i_1 represent variable i at time slice 0 and 1, respectively.

Why do we have the other columns, 'A', 'B', and 'C', that are not attached to particular time slices? What do they represent?

I noticed that the provided code asserts that the given node id can be found (so does require them). dsMap, however, is only updated on the "slice" columns 'A_0', 'A_1', 'B_0', 'B_1', 'C_0', and 'C_1'.

Thank you!
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Re: Smilearn - tutorial 6 - why do we need A,B,C?

Post by shooltz[BayesFusion] »

The tutorial is quite old. The preferred conventions for column names is to use node id only for 0th slice and node id + underscore + slice number for subsequent slices. GeNIe's UI for EM learning uses this convention for automatic matching, including the slices > 0. However, SMILearn's DSL_dataset::MatchNetwork will only match 0th slice based on identical node ids and column names (hence A instead of A_0).
Post Reply