News for hackers. Stuff that matters.
Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by zeratul on Saturday November 05, 2016 @11:22PM (#12883322)
99.3% accuracy is not that great. Currently the best performer on this data set has 99.77% accuracy:

http://yann.lecun.com/exdb/mnist/

Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by paulsutter on Sunday November 06, 2016 @01:20AM (#12883666)
99.3% is pretty good for a one hour lecture that begins with a single-layer network.
Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by danieltillett on Saturday November 05, 2016 @11:36PM (#12883373)
What is the human accuracy on this data set?
Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by mamon on Sunday November 06, 2016 @06:49AM (#12884679)
more interesting question: if some of those "digits" are so hard to recognize even by humans then how can we ever label them with the one true "correct" answer?

Do we know a person who draw those digits and ask "what artist had in mind when making this masterpiece" ? And even then someone might have been trying to draw the "2" but end effect looks more like "3".

I think that some of the test cases simply don't have definitive answer and trying to reach 100% accuracy is just misguided effort.

Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by danieltillett on Sunday November 06, 2016 @07:18AM (#12884772)
Another interesting question is which approach most closely matches the errors made by humans.
Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by gajomi on Saturday November 05, 2016 @11:38PM (#12883377)
I wonder, is there a good reason why these accuracies are reported on a logscale by convention (say, as power of 2 from 50%)?
Re:Tensorflow and Deep Learning, Without a PhD, Martin Gorne... by mamon on Sunday November 06, 2016 @07:00AM (#12884720)
To make even small difference look bigger, cause even a 0.1% improvement over previous best result is considered big deal.