o*********y
发帖数: 65 | 1 Suppose we have a function F(X) that we can evaluate. X is a vector, and
suppose F() is convex. In this case, what is the best strategy (minimal steps
to find minimal) to find the minimal?
My current naive idea needs 2^n+n+1 times evaluation each step (n is the
dimention of X). I guess this should be a classic problem, I wonder what's
the best strategies known? thanks a lot, bow. | h***o
发帖数: 539 | 2 解方程组dF(X)/dX = 0
这就看你的solver有多快了。hoho
不过这样好象不能保证找全所有的minima呀
不如你说说你怎么做的，让我也开开眼界
【在 o*********y 的大作中提到】
: Suppose we have a function F(X) that we can evaluate. X is a vector, and
: suppose F() is convex. In this case, what is the best strategy (minimal steps
: to find minimal) to find the minimal?
: My current naive idea needs 2^n+n+1 times evaluation each step (n is the
: dimention of X). I guess this should be a classic problem, I wonder what's
: the best strategies known? thanks a lot, bow.
| h***o
发帖数: 26 | 3 convex surface, only one minima
newton method should be most effiecient
【在 h***o 的大作中提到】
: 解方程组dF(X)/dX = 0
: 这就看你的solver有多快了。hoho
: 不过这样好象不能保证找全所有的minima呀
: 不如你说说你怎么做的，让我也开开眼界
|
|
