Failure in initial objective function evaluation. FMINUNC cannot continue. Need some help

Question

奥刘 2020-9-22

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/597769-failure-in-initial-objective-function-evaluation-fminunc-cannot-continue-need-some-help

评论：奥刘 2020-9-22

It's my first time to use fminunc solver. R_pq,array_x,array_y,M,N,R00 are constants I have set.

x0 = -pi + (pi+pi)*rand(M,N);
options = optimoptions(@fminunc,'Algorithm','trust-region','SpecifyObjectiveGradient',true);
[x,fval,output]=fminunc(@(x) error_fun(x,R_pq,array_x,array_y,M,N,R00),x0,options);

Without options, the code can function without error. But after I set the options, it shows

Error using error_fun
Too many output arguments.
Error in main_beamshaping_336_XZLZ_AMM_phaseonly>@(x)error_fun(x,R_pq,array_x,array_y,M,N,R00)
(line 128)
[x,fval,output]=fminunc(@(x) error_fun(x,R_pq,array_x,array_y,M,N,R00),x0,options);
Error in fminunc (line 303)
        [f,GRAD] = feval(funfcn{3},x,varargin{:});
Error in main_beamshaping_336_XZLZ_AMM_phaseonly (line 128)
[x,fval,output]=fminunc(@(x) error_fun(x,R_pq,array_x,array_y,M,N,R00),x0,options);
Caused by:
    Failure in initial objective function evaluation. FMINUNC cannot continue.

I am confused about what is wrong. I would be very grateful indeed for any help

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Matt J 2020-9-22

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/597769-failure-in-initial-objective-function-evaluation-fminunc-cannot-continue-need-some-help#answer_498370

编辑：Matt J 2020-9-22

You have set SpecifyObjectiveGradient=true, but your objective function code does not provide a second output for your gradient calculation.

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

奥刘 2020-9-22

OK, thanks! I will follow the instructions below.

请先登录，再进行评论。

Answer 2

Walter Roberson 2020-9-22

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/597769-failure-in-initial-objective-function-evaluation-fminunc-cannot-continue-need-some-help#answer_498373

When you have the option SpecifyObjectiveGradient, then your function must be able to return the gradient as well as the objective value. It should check nargout and only set the second output if nargout > 1

When you use that option, your nonlinear constraint function (if any) should have two extra outputs. See https://www.mathworks.com/help/optim/ug/nonlinear-constraints.html#bskkri8 "Including Gradients in Constraint Functions"