help needed for matlab error levenberg marquardt

i am using matlab 2015 but how to get out of this error?

Error using optimset (line 213)
  Unrecognized parameter name 'LevenbergMarquardt'.  Please see the options table in the documentation for a list of
  acceptable option parameters. Note that some parameters are only supported by OPTIMOPTIONS. Link to options table

    Error in gluc_mm_mle (line 83)
    options = optimset('Display','iter','TolFun', 1e-4,...%'iter' default:1e-4

Answer 1

[Star Strider]

To begin with, you’re mis-spelling it. From the documentation (for optimoptions):

options = optimoptions(@lsqnonlin,'Algorithm','levenberg-marquardt','MaxFunEvals',1500)

That still may not work, depending on what solver you’re using.

In order to find out what options you have, again from the documentation for optimset:

• options = optimset(optimfun) creates an options structure options with all option names and default values relevant to the optimization function optimfun.

[divya]

my partial code is

options = optimset('Display','iter','TolFun', 1e-4,...%default: 1e-4
'TolX',1e-5,... %default: 1e-4
'LevenbergMarquardt','on',... %default: on
'LargeScale','on'); %default: on
plt = 0;
%LSQNONLIN: objective function should return the model error
[p_est,resnorm,RESIDUAL,exitflag,OUTPUT,LAMBDA,Jacobian] =lsqnonlin(@obj_fn,p_init,lb,ub,options, p_fix,gluc_exp,tspan,tu, plt); disp(' ')
%Accuracy:
%lsqnonlin returns the Jacobian as a sparse matrix
varp = resnorm*inv(Jacobian'*Jacobian)/length(tspan);
stdp = sqrt(diag(varp)); %The standard deviation is the square root
%of the variance
%p = [Sg, Gb, k3, Si, Ib];
p = [p_est(1), p_fix(1), p_est(2), p_est(3), p_fix(2)];
%x0 = [G0, X0]
x0 = [p_est(4), p_fix(3)];
disp(' Parameters:')
disp([' Sg = ', num2str(p_est(1)), ' +/- ', num2str(stdp(1))])
disp([' Gb = ', num2str(p_fix(1))])
disp([' k3 = ', num2str(p_est(2)), ' +/- ', num2str(stdp(2))])
disp([' Si = ', num2str(p_est(3)), ' +/- ', num2str(stdp(3))])
disp([' Ib = ', num2str(p_fix(2))])
disp(' Initial conditions:')
disp([' G0 = ', num2str(p_est(4)), ' +/- ', num2str(stdp(4))])
disp([' X0 = ', num2str(p_fix(3))]); disp(' ')
plt = 1;
gluc = gluc_sim(tspan,x0,tu, p,plt);
%compare model output with measured data
figure(2); subplot(221)
h1 = plot(t_gluc,gluc_exp,'or', 'Linewidth',2);
%legend(h1, 'experimental test data')
still same error

Error using optimset (line 213) Unrecognized parameter name 'LevenbergMarquardt'. Please see the options table in the documentation for a list of acceptable option parameters. Note that some parameters are only supported by OPTIMOPTIONS. Link to options table

Error in gluc_mm_mle (line 83)
options = optimset('Display','iter','TolFun', 1e-4,...%'iter' default:1e-4

[Star Strider]

You cannot define your optimset call with comments, so do not to this:

options = optimset('Display','iter','TolFun', 1e-4,...%default: 1e-4
'TolX',1e-5,... %default: 1e-4
'LevenbergMarquardt','on',... %default: on
'LargeScale','on'); %default: on

Try this instead:

options = optimset('Display','iter','TolFun', 1e-4,...
'TolX',1e-5,... 
'LevenbergMarquardt','on',... 
'LargeScale','on');

However, this may not be your only problem. If you are using the lsqnonlin function, you can most easily determine what options are available to you with:

options = optimset(@lsqnonlin)

Note that the algorithm specification is part of a name-value pair. You have to include it as: 'Algorithm','levenberg-marquardt' with those exact spellings. MATLAB will not recognise alternative spellings in its arguments.

So your options structure would have to be restated as:

options = optimset('Display','iter','TolFun', 1e-4,...
'TolX',1e-5,... 
'Algorithm','levenberg-marquardt',... 
'LargeScale','on');

This should work if the other parts of your optimset call are correct.

I also don’t know what version of MATLAB you’re using. The current online documentation (and the version I’m using) are R2015b, and the documentation for one version does not always apply to other versions.

[divya]

thank you... for the reply.i have tried the above code and after re correcting i am getting the below error. i am using matlab2015b

Error in lsqnonlin (line 194)
            initVals.F = feval(funfcn{3},xCurrent,varargin{:});

Error in gluc_mm_mle (line 85)
[p_est,resnorm,Residual,exitflag,output,lambda,Jacobian] =lsqnonlin(@obj_fn,p_init,lb,ub,options,...

Caused by:
    Failure in initial user-supplied objective function evaluation. LSQNONLIN cannot continue

[Star Strider]

For some unexplained reason, your objective function is failing at the outset. From the code you posted, I have no idea what the problem could be.

My guess is that you’re not passing the extra parameters correctly. If ‘B’ is the parameter vector you’re solving for, perhaps you need to do something like this:

function y = obj_fn(B,p_fix,gluc_exp,tspan,tu, plt)
      . . .
end

then call it in lsqnonlin as:

[p_est,resnorm,RESIDUAL,exitflag,OUTPUT,LAMBDA,Jacobian] =lsqnonlin(@(B)obj_fn(B,p_fix,gluc_exp,tspan,tu, plt), p_init,lb,ub,options);

Some ways of passing parameters are better than others, and earlier methods could be obsolete. You will have to adapt my idea to your code, but something like it should work.

[divya]

i have used this file....

function e = obj_fn(p_var, p_fix,data,tspan,tu, sigma_nu,sigma_mu,plt)
  %4 unknown model parameters:
  % p_var = [Sg k3 Si G0];
  %3 known parameters:
  % p_fix = [Gb Ib x0(2)];
  %p = [Sg, Gb, k3, Si, Ib];
  p = [p_var(1), p_fix(1), p_var(2), p_var(3), p_fix(2)];
  %x0 = [G0, X0]
  x0 = [p_var(4), p_fix(3)];
  gluc = *gluc_sim*(tspan,x0,tu, p, sigma_nu,sigma_mu,0);
  %LSQNONLIN: objective function should return the model error
  e = gluc-data;
  if plt==1 %fast update during estimation process
   N=length(tspan);
   figure(2)
   plot(1:N,gluc,'b',1:N,data,'r'); drawnow
  end

and for glu_sim

function gluc = gluc_sim(tspan,x0,tu, p,plt)
%SIM_INPUT_SIM Simulation of glucose minimal model.
%x0: initial conditions for glucose conc. and state variable denoting
% insulin action
%gluc: model output, glucose conc. in plasma [mg/dL]
if tspan(1) < tu(1,1)
error(' no input defined at start of simulation (compare tspan(1)) vs. tu(1,1)')
end
ode_options = [];
[t,x] = ode45(@gluc_ode,tspan,x0,ode_options, tu, p);
%Output
gluc = x(:,1);
if plt==1
 gluc_plt(tspan,x,tu,p,sigma_nu,sigma_mu)
end

[Walter Roberson]

Star Strider: FYI, the syntax

options = optimset('Display','iter','TolFun', 1e-4,...%default: 1e-4
'TolX',1e-5,... %default: 1e-4
'LevenbergMarquardt','on',... %default: on
'LargeScale','on'); %default: on

is fine. The ... operator defines everything remaining on the line as being a comment to be ignored, so it is as if those %default things are not there at all as far as MATLAB is concerned.

[Star Strider]

I cannot make any sense of that.

This is invalid MATLAB syntax:

gluc = *gluc_sim*(tspan,x0,tu, p, sigma_nu,sigma_mu,0);