自己顶一个，没有人知道的吗？期望高手出现

我顶，希望有高手能帮助一下。

function [x,FVAL,EXITFLAG,OUTPUT,JACOB] = fsolve(FUN,x,options,varargin)
%FSOLVE solves systems of nonlinear equations of several variables.
%
%   FSOLVE attempts to solve equations of the form:
%            
%   F(X)=0    where F and X may be vectors or matrices.   
%
%   X=FSOLVE(FUN,X0) starts at the matrix X0 and tries to solve the
%   equations in FUN.  FUN accepts input X and returns a vector (matrix) of
%   equation values F evaluated at X.
%
%   X=FSOLVE(FUN,X0,OPTIONS) solves the equations with the default optimization
%   parameters replaced by values in the structure OPTIONS, an argument
%   created with the OPTIMSET function.  See OPTIMSET for details.  Used
%   options are Display, TolX, TolFun, DerivativeCheck, Diagnostics,
%   FunValCheck, Jacobian, JacobMult, JacobPattern, LineSearchType,
%   NonlEqnAlgorithm, MaxFunEvals, MaxIter, OutputFcn, DiffMinChange and
%   DiffMaxChange, LargeScale, MaxPCGIter, PrecondBandWidth, TolPCG, and
%   TypicalX. Use the Jacobian option to specify that FUN also returns a
%   second output argument J that is the Jacobian matrix at the point X. If
%   FUN returns a vector F of m components when X has length n, then J is
%   an m-by-n matrix where J(i,j) is the partial derivative of F(i) with
%   respect to x(j). (Note that the Jacobian J is the transpose of the
%   gradient of F.)
%
%   [X,FVAL]=FSOLVE(FUN,X0,...) returns the value of the equations FUN at X.
%
%   [X,FVAL,EXITFLAG]=FSOLVE(FUN,X0,...) returns an EXITFLAG that describes the
%   exit condition of FSOLVE. Possible values of EXITFLAG and the corresponding
%   exit conditions are
%
%     1  FSOLVE converged to a solution X.
%     2  Change in X smaller than the specified tolerance.
%     3  Change in the residual smaller than the specified tolerance.
%     4  Magnitude of search direction smaller than the specified tolerance.
%     0  Maximum number of function evaluations or iterations reached.
%    -1  Algorithm terminated by the output function.
%    -2  Algorithm seems to be converging to a point that is not a root.
%    -3  Trust region radius became too small.
%    -4  Line search cannot sufficiently decrease the residual along the current
%         search direction.
%
%   [X,FVAL,EXITFLAG,OUTPUT]=FSOLVE(FUN,X0,...) returns a structure OUTPUT
%   with the number of iterations taken in OUTPUT.iterations, the number of
%   function evaluations in OUTPUT.funcCount, the algorithm used in OUTPUT.algorithm,
%   the number of CG iterations (if used) in OUTPUT.cgiterations, the first-order
%   optimality (if used) in OUTPUT.firstorderopt, and the exit message in
%   OUTPUT.message.
%
%   [X,FVAL,EXITFLAG,OUTPUT,JACOB]=FSOLVE(FUN,X0,...) returns the
%   Jacobian of FUN at X.  
%
%   Examples
%     FUN can be specified using @:
%        x = fsolve(@myfun,[2 3 4],optimset('Display','iter'))
%
%   where myfun is a MATLAB function such as:
%
%       function F = myfun(x)
%       F = sin(x);
%
%   FUN can also be an anonymous function:
%
%       x = fsolve(@(x) sin(3*x),[1 4],optimset('Display','off'))
%
%   If FUN is parameterized, you can use anonymous functions to capture the
%   problem-dependent parameters. Suppose you want to solve the system of
%   nonlinear equations given in the function myfun, which is parameterized
%   by its second argument c. Here myfun is an M-file function such as
%     
%       function F = myfun(x,c)
%       F = [ 2*x(1) - x(2) - exp(c*x(1))
%             -x(1) + 2*x(2) - exp(c*x(2))];
%           
%   To solve the system of equations for a specific value of c, first assign the
%   value to c. Then create a one-argument anonymous function that captures
%   that value of c and calls myfun with two arguments. Finally, pass this anonymous
%   function to FSOLVE:
%
%       c = -1; % define parameter first
%       x = fsolve(@(x) myfun(x,c),[-5;-5])
%
%   See also OPTIMSET, LSQNONLIN, @, INLINE.

%   Copyright 1990-2004 The MathWorks, Inc.
%   $Revision: 1.41.4.8 $  $Date: 2004/12/24 20:46:40 $

% ------------Initialization----------------

defaultopt = struct('Display','final','LargeScale','off',...
   'NonlEqnAlgorithm','dogleg',...
   'TolX',1e-6,'TolFun',1e-6,'DerivativeCheck','off',...
   'Diagnostics','off','FunValCheck','off',...
   'Jacobian','off','JacobMult',[],...% JacobMult set to [] by default
   'JacobPattern','sparse(ones(Jrows,Jcols))',...
   'MaxFunEvals','100*numberOfVariables',...
   'DiffMaxChange',1e-1,'DiffMinChange',1e-8,...
   'PrecondBandWidth',0,'TypicalX','ones(numberOfVariables,1)',...
   'MaxPCGIter','max(1,floor(numberOfVariables/2))', ...
   'TolPCG',0.1,'MaxIter',400,...
   'LineSearchType','quadcubic','LevenbergMarquardt','off', ...
   'OutputFcn', []);

% If just 'defaults' passed in, return the default options in X
if nargin==1 && nargout <= 1 && isequal(FUN,'defaults')
   x = defaultopt;
   return
end

if nargin < 2
  error('optim:fsolve:NotEnoughInputs','FSOLVE requires two input arguments.')
end
if nargin < 3, options=[]; end

% Check for non-double inputs
if ~isa(x,'double')
  error('optim:fsolve:NonDoubleInput', ...
        'FSOLVE only accepts inputs of data type double.')
end

% These are added so that we can have the same code as in lsqnonlin which
%  actually has upper and lower bounds.
LB = []; UB = [];

xstart=x(;
numberOfVariables=length(xstart);

large        = 'large-scale';
medium       = 'medium-scale: line search';
dogleg       = 'trust-region dogleg';

switch optimget(options,'Display',defaultopt,'fast')
    case {'off','none'}
        verbosity = 0;
    case 'iter'
        verbosity = 2;
    case 'final'
        verbosity = 1;
    case 'testing'
        verbosity = Inf;
    otherwise
        verbosity = 1;
end
diagnostics = isequal(optimget(options,'Diagnostics',defaultopt,'fast'),'on');
gradflag =  strcmp(optimget(options,'Jacobian',defaultopt,'fast'),'on');
% 0 means large-scale trust-region, 1 means medium-scale algorithm
mediumflag = strcmp(optimget(options,'LargeScale',defaultopt,'fast'),'off');
funValCheck = strcmp(optimget(options,'FunValCheck',defaultopt,'fast'),'on');
switch optimget(options,'NonlEqnAlgorithm',defaultopt,'fast')
    case 'dogleg'
        algorithmflag = 1;
    case 'lm'
        algorithmflag = 2;
    case 'gn'
        algorithmflag = 3;
    otherwise
        algorithmflag = 1;
end
mtxmpy = optimget(options,'JacobMult',defaultopt,'fast');
if isequal(mtxmpy,'atamult')
    warning('optim:fsolve:NameClash', ...
        ['Potential function name clash with a Toolbox helper function:\n' ...
        'Use a name besides ''atamult'' for your JacobMult function to\n' ...
        'avoid errors or unexpected results.'])
end

% Convert to inline function as needed
if ~isempty(FUN)  % will detect empty string, empty matrix, empty cell array
    funfcn = lsqfcnchk(FUN,'fsolve',length(varargin),funValCheck,gradflag);
else
    error('optim:fsolve:InvalidFUN', ...
        ['FUN must be a function name, valid string expression, or inline object;\n' ...
        ' or, FUN may be a cell array that contains these type of objects.'])
end

JAC = [];
x( = xstart;
switch funfcn{1}
    case 'fun'
        fuser = feval(funfcn{3},x,varargin{:});
        f = fuser(;
        nfun=length(f);
    case 'fungrad'
        [fuser,JAC] = feval(funfcn{3},x,varargin{:});
        f = fuser(;
        nfun=length(f);
    case 'fun_then_grad'
        fuser = feval(funfcn{3},x,varargin{:});
        f = fuser(;
        JAC = feval(funfcn{4},x,varargin{:});
        nfun=length(f);
    otherwise
        error('optim:fsolve:UndefinedCalltype','Undefined calltype in FSOLVE.')
end

if gradflag
    % check size of JAC
    [Jrows, Jcols]=size(JAC);
    if isempty(mtxmpy)
        % Not using 'JacobMult' so Jacobian must be correct size
        if Jrows~=nfun || Jcols~=numberOfVariables
            error('optim:fsolve:InvalidJacobian', ...
                ['User-defined Jacobian is not the correct size:\n' ...
                ' the Jacobian matrix should be %d-by-%d.'],nfun,numberOfVariables)
        end
    end
else
    Jrows = nfun;
    Jcols = numberOfVariables;
end

YDATA = []; caller = 'fsolve';

% large-scale method and enough equations (as many as variables)
if ~mediumflag && nfun >= numberOfVariables
    OUTPUT.algorithm = large;

    % large-scale method and not enough equations --
    % switch to medium-scale algorithm
elseif ~mediumflag && nfun < numberOfVariables
    warning('optim:fsolve:FewerFunsThanVars', ...
        ['Large-scale method requires at least as many equations as variables;\n' ...
        ' switching to line-search method instead.'])
    OUTPUT.algorithm = medium;

    % medium-scale and no bounds
elseif mediumflag && isempty(LB) && isempty(UB)
    if algorithmflag == 1 && nfun == numberOfVariables % dogleg method
        OUTPUT.algorithm = dogleg;
    else
        if algorithmflag == 1 && nfun ~= numberOfVariables
            warning('optim:fsolve:NonSquareSystem', ...
                ['Default trust-region dogleg method of FSOLVE cannot\n handle non-square systems; ', ...
                'switching to Gauss-Newton method.']);
            algorithmflag = 3;
        end
        OUTPUT.algorithm = medium;
        if algorithmflag == 2
            % Calling nlsq which looks at LevenbergMarquardt option
            % (changing option "unsafely" for speed; users should use optimset)
            options.LevenbergMarquardt = 'on';
        else % algorithmflag == 3
            % (changing option "unsafely" for speed; users should use optimset)
            options.LevenbergMarquardt = 'off';
        end
    end

    % medium-scale and bounds and enough equations, switch to trust region
elseif mediumflag && (~isempty(LB) || ~isempty(UB)) && nfun >= numberOfVariables
    warning('optim:fsolve:LineSearchAndBounds', ...
        ['Line-search method does not handle bound constraints;\n' ...
        ' switching to large-scale trust-region method instead.'])
    OUTPUT.algorithm = large;

    % can't handle this one:
elseif mediumflag && (~isempty(LB) || ~isempty(UB)) && nfun < numberOfVariables
    error('optim:fsolveroblemNotHandled', ...
        ['Line-search method does not handle bound constraints and trust-region\n' ...
        ' method requires at least as many equations as variables; aborting.'])
end

if diagnostics > 0
    % Do diagnostics on information so far
    constflag = 0; gradconstflag = 0; non_eq=0;non_ineq=0;lin_eq=0;lin_ineq=0;
    confcn{1}=[];c=[];ceq=[];cGRAD=[];ceqGRAD=[];
    hessflag = 0; HESS=[];
    diagnose('fsolve',OUTPUT,gradflag,hessflag,constflag,gradconstflag,...
        mediumflag,options,defaultopt,xstart,non_eq,...
        non_ineq,lin_eq,lin_ineq,LB,UB,funfcn,confcn,f,JAC,HESS,c,ceq,cGRAD,ceqGRAD);

end

% Execute algorithm
if isequal(OUTPUT.algorithm, large)
    if ~gradflag
        Jstr = optimget(options,'JacobPattern',defaultopt,'fast');
        if ischar(Jstr)
            if isequal(lower(Jstr),'sparse(ones(jrows,jcols))')
                Jstr = sparse(ones(Jrows,Jcols));
            else
                error('optim:fsolve:InvalidJacobPattern', ...
                    'Option ''JacobPattern'' must be a matrix if not the default.')
            end
        end
    else
        Jstr = [];
    end
    computeLambda = 0;
    [x,FVAL,LAMBDA,JACOB,EXITFLAG,OUTPUT,msg]=...
        snls(funfcn,x,LB,UB,verbosity,options,defaultopt,f,JAC,YDATA,caller,...
        Jstr,computeLambda,varargin{:});
elseif isequal(OUTPUT.algorithm, dogleg)
    % trust region dogleg method
    Jstr = [];
    [x,FVAL,JACOB,EXITFLAG,OUTPUT,msg]=...
        trustnleqn(funfcn,x,verbosity,gradflag,options,defaultopt,f,JAC,...
        YDATA,Jstr,varargin{:});
else
    % line search (Gauss-Newton or Levenberg-Marquardt)
    [x,FVAL,JACOB,EXITFLAG,OUTPUT,msg] = ...
        nlsq(funfcn,x,verbosity,options,defaultopt,f,JAC,YDATA,caller,varargin{:});
end

Resnorm = FVAL'*FVAL;  % assumes FVAL still a vector
if EXITFLAG > 0 % if we think we converged:
    if Resnorm > sqrt(optimget(options,'TolFun',defaultopt,'fast'))
        OUTPUT.message = ...
            sprintf(['Optimizer appears to be converging to a minimum that is not a root:\n' ...
            'Sum of squares of the function values is > sqrt(options.TolFun).\n' ...
            'Try again with a new starting point.']);
        if verbosity > 0
            disp(OUTPUT.message)
        end
        EXITFLAG = -2;
    else
        OUTPUT.message = msg;
        if verbosity > 0
            disp(OUTPUT.message);
        end
    end
else
    OUTPUT.message = msg;
    if verbosity > 0
        disp(OUTPUT.message);
    end
end

% Reset FVAL to shape of the user-function output, fuser
FVAL = reshape(FVAL,size(fuser));，

好像用的是二分法逐步逼近
我不懂什么上山下山的

不是很懂。。。

如果有什么不懂得,欢迎到计算模拟版块交流

上面的就是源程序