News for hackers. Stuff that matters.
Re:Machine learning works spectacularly well, but mathematic... by robotresearcher on Friday December 04, 2015 @01:22PM (#10678027)
Because the solution space is convex if you've chosen your representation well.
Re:Machine learning works spectacularly well, but mathematic... by gipp on Friday December 04, 2015 @02:41PM (#10678582)
This is definitely not the case for general neural networks, though.
Re:Machine learning works spectacularly well, but mathematic... by robotresearcher on Friday December 04, 2015 @03:22PM (#10678883)
If gradient descent is working reliably, the problem is convex. See the sibling comments for the intuition for large dimensional spaces.
Re:Machine learning works spectacularly well, but mathematic... by gipp on Saturday December 05, 2015 @09:48AM (#10682000)
"The problem is convex" and "the algorithm is unlikely to get stuck in a local minimum on realistic problems" are very different things.
Re:Machine learning works spectacularly well, but mathematic... by robotresearcher on Saturday December 05, 2015 @02:32PM (#10682982)
Right. Hence the word 'reliably'.