New working notes: "Reduced MNIST: how well can machines learn from small data?" http://cognitivemedium.com/rmnist
Show this thread
-
Their result of RMNIST/10: 92.07% is not bad at all.
I'd be surprised if it's not possible to get above 99%. But it might take quite some iteration to get there!
-
-
That's better than a few semi-supervised techniques I know of. To be fair those techniques don't use convnets with dropout and augmentation but still quite interesting.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Philosophical question: I'm sure we can get there, but at what point are we overfitting the technique to the problem space?
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.