Training and testing partitions

17 views
Skip to first unread message

Hossam

unread,
Mar 20, 2012, 3:10:29 PM3/20/12
to HeuristicLab
Hi All,

I have noticed that when I start a new symbolic regression problem,
the data is divided by default 0 to 12 for training and 12 to 25 for
testing which means there are 26 rows. but when I view the Dataset I
can see only 25 rows. Its the same when I load my own Dataset. Should
I modify the starting index to 1 instead of 0. In the default case, is
the row 12 included in both training and testing ?
Please may you explain the indexing of the Dataset ?

Gabriel Kronberger

unread,
Mar 21, 2012, 4:15:12 AM3/21/12
to heuris...@googlegroups.com
Dear Hossam,

in HeuristicLab we use the convention that the start index of a range is
inclusive and the end index is exclusive.
The training partition contains rows [0..12[ and the test partition contains
rows [12..25[. This represents a 50/50 split of the available rows which is
also used as a default when loading new data into a symbolic regression
problem. However, you can change the start and end indexes as you wish.

Best,
Gabriel

Reply all
Reply to author
Forward
0 new messages