News for hackers. Stuff that matters.
Re:Pruning AI networks without impacting performance by sushisource on Tuesday December 12, 2017 @04:59PM (#15909914)
Neat. As someone who has largely surface-level knowledge of ML, does this development mean we can expect many more on-phone networks in the future?

Phrased another way: Is this a huge step up from previous pruning methods?

Re:Pruning AI networks without impacting performance by CardenB on Tuesday December 12, 2017 @07:07PM (#15910763)
Glancing at the paper, I see the biggest dataset they used was MNIST. So it's hard to quantify how much error they preserve for larger, more useful networks in more complicated tasks.