News for hackers. Stuff that matters.

A comment on Google A.I. researchers develop alternative architecture for neural networks. Parent comment

Re:Google A.I. researchers develop alternative architecture ... by zardo on Friday November 10, 2017 @01:39PM (#15671910)
>just brute-forcing, trial-and-error.

I think as a whole, the community is executing a distributed epsilon-greedy montecarlo search, which is theoretically guaranteed to converge on an optimal policy eventually.

[ Parent | on HN ]
Re:Google A.I. researchers develop alternative architecture ... by inverse_pi on Friday November 10, 2017 @05:05PM (#15673445)
not necessarily. The convergence point can be and will probably be sub-optimal if we keep doing this without a mathematical framework guiding us.
[ Parent | on HN ]
Re:Google A.I. researchers develop alternative architecture ... by zardo on Friday November 10, 2017 @05:13PM (#15673505)
So long as you never stop trying random things, you are guaranteed to find a global optimum... as time->inf

Not a very useful guarantee, but that's theoretical guarantees for you.

[ Parent | on HN ]
Re:Google A.I. researchers develop alternative architecture ... by goptimize on Friday November 10, 2017 @02:26PM (#15672263)
when time goes to infinity which is a shitty guarantee
[ Parent | on HN ]