someone tried to convince me that held out data wasn't useful 
It's well-known that you're able to report higher metrics to your boss when you evaluate your model on the training data. Clearly validation data only gets in the way of 100% accuracy
-
-
They did at least follow it up with "but I've got no experience building a real model so I might be wrong"... Might be...
-
Related problem: only testing ML models via downstream impact on live data via AB tests. Has anyone written about the pitfalls of this?
- Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.