MATLAB: How to use fminunc providing gradient and hessian

fminconfminuncoptimization

Hi, could you help me in making this code (attached) working? Thanks!

Best Answer

You need to return the gradient and Hessian

function [val,g,H]=log_lik(theta,ix)

Also, these tolerances look unrealistically small

'TolX',10^-30,'TolFun',10^-30

Related Solutions

MATLAB: Calculate the gradient of a function

You are confusing the numeric gradient function http://www.mathworks.com/help/matlab/ref/gradient.html with the symbolic gradient function, http://www.mathworks.com/help/symbolic/gradient.html

On the other hand, neither gradient() accepts a vector or cell array of function handles. Numeric gradient() accepts a numeric vector or array, and spacing distances for each of the dimensions. Symbolic gradient() accepts a scalar symbolic expression or symbolic function together with the variables to take the gradient over.

When you have a vector of functions to work with, you almost certainly want a jacobian rather than a gradient. Symbolic jacobian http://www.mathworks.com/help/symbolic/jacobian.html accepts a symbolic expression, symbolic function, or symbolic vector (but not a cell array of function handles.)

MATLAB: Fminunc: get Hessian approximation after each iteration

You could try running fminunc 1 iteration at a time in a for-loop and saving the 6th output arguments,


  options.MaxIter=1;
  for i=1:N
     
   [x,~,~,~,~,hessian{i}] = fminunc(fun,x,options);
  end

Best Answer

Related Solutions

MATLAB: Calculate the gradient of a function

MATLAB: Fminunc: get Hessian approximation after each iteration

Related Question