01-17-2017 09:48 AM
I'm strictly a newbie, and i've been experimenting for a while, mostly with the new Deep Learning module. So far so good, but recently i noticed that the (regression) performance module gives me performance-values that are not the same as the values given by H2O. I do remember that at an earlier moment, this was not the case, but i can't figure out what has changed. The H2O values seem more reliable Any ideas?
I've added an example of a process
Solved! Go to Solution.
01-17-2017 10:10 AM
the performance measures in the description of DL are training errors, not real prediction errors. Thus they are different to your splitted data and also should be treated with care. We always recommend to use Cross validation.