Dear SMILE Team,
Here is another variation of a network that has odd information implications.
We have a link from C-> B, but this link is uninformative. The outcome of C does not affect the outcome of B so observing B does not tell us about C.
However, observing B still tells us about A. Unfortunately, the test view of the diagnosis system reports that B has zero information value.
A->B
C->{B,D}
Regards,
Bob.
Oddities in information gain calculation
-
- Posts: 5
- Joined: Wed Sep 23, 2009 2:00 am
Oddities in information gain calculation
- Attachments
-
- simple_dx_net3.xdsl
- (2.01 KiB) Downloaded 333 times
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Oddities in information gain calculation
Again, this is caused by default selection of the pursued fault. I guess you should pursue both A:FAIL and C:FAIL.