-
Hi, |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 4 replies
-
How big are the training and validation sets? Are you maybe using a |
Beta Was this translation helpful? Give feedback.
-
Yes. See here, for example (assuming you're using RV via CLI by defining a
This is a more general ML question (i.e. not RV-specific) and I don't think there is a simple answer to this. You might want to consider things like cross-validation. |
Beta Was this translation helpful? Give feedback.
In that case, the variance makes sense because of the small size of the validation set. One option might be to train for longer, in the hopes that the scores converge.
Re: the sliding windows: it is best to use sliding windows for the validation set so that it is identical across experiments, which makes comparison of metrics more meaningful.