3 ms·
Why can't we check for overfitting using the usual ML techniques?
by rozab 7y ago
Why can't we check for overfitting using the usual ML techniques?
- naniwaduni 7y agoWe know the tests are overfitting. That's not really a problem we can worry about too much; the tests can't be perfectly comprehensive and also finite. The problem is that we're using to use the model to tell us something about adversarial input.