朱莉娅:最小化带有多个参数的函数(BFGS) [英] Julia: Minimise a function with multiple arguments (BFGS)

问题描述

我正在尝试使用BFGS算法在Optim.jl库中最小化具有多个参数的函数.

I am trying to minimise a function with multiple arguments with the Optim.jl library, using a BFGS algorithm.

在Optim库的GitHub网站上，我找到了以下工作示例:

In the GitHub website of the Optim library, I found the following working example:

using Optim
rosenbrock(x) = (1.0 - x[1])^2 + 100.0 * (x[2] - x[1]^2)^2
result        = optimize(rosenbrock, zeros(2), BFGS())

假设我的目标函数是:

fmin(x, a) = (1.0 - x[1])^a + 100.0 * (x[2] - x[1]^2)^(1-a)

如何使用 optimize 传递附加的-常数-参数 a ?

How can I pass the additional - constant - argument a using optimize?

推荐答案

最简单的方法是传递一个变量的匿名函数，该变量使用参数集调用原始函数.例如，使用fmin的变体:

The simplest way is to pass an anonymous function of one variable which calls your original function with the parameters set. For example, using a variant of your fmin:

julia> fmin(x, a) = (1.0 - x[1])^a + 100.0 * (x[2] - x[1]^2)^(a)
fmin (generic function with 1 method)

julia> r = optimize(x->fmin(x, 2), zeros(2), BFGS());

julia> r.minimizer, r.minimum
([1.0,1.0],5.471432684244042e-17)

或者，您可以为一个变量创建一个单独的命名函数，该函数将关闭您喜欢的任何参数.没有与Python中scipy.optimize.minimize中的args等效的功能，在Python中，您将不变的参数分别作为元组AFAIK传递.

Alternatively you can make a separate named function of one variable which closes over whatever parameters you like. There's no equivalent to the args in scipy.optimize.minimize in Python where you pass the non-varying arguments separately as a tuple, AFAIK.

这篇关于朱莉娅:最小化带有多个参数的函数(BFGS)的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！