Exciting paper from former colleagues: a lot of noisy data > a little quality data. Head of ML at large autonomous vehicle team told me yesterday they found the exact *opposite*. Interested to compare both! https://arxiv.org/abs/1712.04577 @zacharylipton @AnimaAnandkumar
-
If noise is inherent to the system, it will be learnt and useful. If noise is due to human error / not intrinsic it leads to the wrong model when above a threshold. 2/2
We learn the noise model for both human error in labeling and inherent ambiguity. Despite in the paper
-
-
Yes it was not a point against it. My thoughts were that the ‘>’ sign needs to be qualified when someone compares noisy and quality data.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.