X1swqazx
2013-04-13, 16:28
老师给的一个m文件
function [StepSize info perf] = ...
search(ObjFun, Point, Step, Rule, varargin)
%Find the answer to problem P: a = argmin f(x+a*d)
%
% This function will use exact or inexact line search to solve the problem.
% When doing inexact line search, you can choose Armijo-Goldstein, Wolfe,
% strong Wolfe or you own criterion.
% Record g as the gradient of f.
%
% Call
% [StepSize err] = bolinesearch(ObjFun, Point, Step)
% [StepSize err] = bolinesearch(ObjFun, Point, Step, Rule)
% [StepSize err] = bolinesearch(ObjFun, Point, Step, Rule, p1,p2,...)
% [StepSize err perf] = bolinesearch(......)
%
% Input
% ObjFun: f & g in P, function_handle
% The function should be declared as a M-function
% [F g] = f(x, p1,p2,...)
% F is a scalar, along with g an n-vector.
% Point: x in P, n-vector
% Step: d in P, n-vector
% Rule: option & method & criterion to solve P, struct
% Rule.crtr: criterion, function_handle
% The function should be declared as a M-function
% Judge = criterion(Step, StepSize, Data0, Data, flag)
% Judge is a logical number with 1 perfect, and Step is just
% as metioned above StepSize below, and Data0 & Data are
% structs including possible fields F g, flag is a array
% including several parameters in the criterion.
% choice: boarmgld, bowlf, bostwlf
% Rule.mthd: method to get new point, function_handle
% The function shoule be declared as a M-function
% [NewSize New] = method(ObjFun, Point, Step, StepSize,
% Data, p1,p2,...)
% notice here StepSize & NewSize can be a scalar array(to
% make coding easy), but only the 1st element is just needed
% actually. And in that situation the Data & New will be
% struct arrays.
% choice: bointrplt22, bointrplt33
% Rule.opt: options of iteration, scalar array
% opt(1): 0 - exact line search, use .618 method
% else - inexact, need crtr & mthd
% (2): upper bound of a
% (3): maximum of iterations
% (4:5): criterion flag, but also can be .618 method flag which
% have one element e with default 1e-3
% default: bostwlf - bointrplt33 - [1 10 10 0.95 0.05]
% or [0 10 25 1e-4] (if opt(1) = 0)
%
请问在使用函数时输入的参数格式是怎样的，又是函数句柄又是结构体我总是输不对
function [StepSize info perf] = ...
search(ObjFun, Point, Step, Rule, varargin)
%Find the answer to problem P: a = argmin f(x+a*d)
%
% This function will use exact or inexact line search to solve the problem.
% When doing inexact line search, you can choose Armijo-Goldstein, Wolfe,
% strong Wolfe or you own criterion.
% Record g as the gradient of f.
%
% Call
% [StepSize err] = bolinesearch(ObjFun, Point, Step)
% [StepSize err] = bolinesearch(ObjFun, Point, Step, Rule)
% [StepSize err] = bolinesearch(ObjFun, Point, Step, Rule, p1,p2,...)
% [StepSize err perf] = bolinesearch(......)
%
% Input
% ObjFun: f & g in P, function_handle
% The function should be declared as a M-function
% [F g] = f(x, p1,p2,...)
% F is a scalar, along with g an n-vector.
% Point: x in P, n-vector
% Step: d in P, n-vector
% Rule: option & method & criterion to solve P, struct
% Rule.crtr: criterion, function_handle
% The function should be declared as a M-function
% Judge = criterion(Step, StepSize, Data0, Data, flag)
% Judge is a logical number with 1 perfect, and Step is just
% as metioned above StepSize below, and Data0 & Data are
% structs including possible fields F g, flag is a array
% including several parameters in the criterion.
% choice: boarmgld, bowlf, bostwlf
% Rule.mthd: method to get new point, function_handle
% The function shoule be declared as a M-function
% [NewSize New] = method(ObjFun, Point, Step, StepSize,
% Data, p1,p2,...)
% notice here StepSize & NewSize can be a scalar array(to
% make coding easy), but only the 1st element is just needed
% actually. And in that situation the Data & New will be
% struct arrays.
% choice: bointrplt22, bointrplt33
% Rule.opt: options of iteration, scalar array
% opt(1): 0 - exact line search, use .618 method
% else - inexact, need crtr & mthd
% (2): upper bound of a
% (3): maximum of iterations
% (4:5): criterion flag, but also can be .618 method flag which
% have one element e with default 1e-3
% default: bostwlf - bointrplt33 - [1 10 10 0.95 0.05]
% or [0 10 25 1e-4] (if opt(1) = 0)
%
请问在使用函数时输入的参数格式是怎样的，又是函数句柄又是结构体我总是输不对