contradicted by the fact that it generalizes surprisingly well on unseen data?
The amount of shit an NN can memorize in its parameters, and how dumb this makes it given unexpected data, dwarfs the idiocy of legacy AI/ML
-
-
-
Depends on the unseen data. NNs memorize much more irrelevant things found in some, but not all unseen data than legacy classifiers.
End of conversation
New conversation -
-
-
The field's term of art for memorizing shit is "overtraining". It doesn't always happen, but it's definitely a serious hazard.
-
I thought it was "overfitting" which in its trivial forms is avoided by having separate train & test sets, but when both are "biased", tough
- 1 more reply
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.