how to pass variables into the cost function and gradient function while using store

[Poo Uoo]

hello, 

I am trying to run the first example using the store. 

But how to pass the constant matrix A into the cost function and gradient function?

I don't want to make the variable A's scope span across multiple functions

% Generate random problem data.

n = 1000;

A = randn(n);

A = .5*(A+A.');

% Create the problem structure.

manifold = spherefactory(n);

problem.M = manifold;

% Define the problem cost function and its Euclidean gradient.

problem.cost = @mycost;

problem.egrad = @myegrad; % notice the 'e' in 'egrad' for Euclidean

% Numerically check gradient consistency (optional).

checkgradient(problem);

% Solve.

[x, xcost, info, options] = trustregions(problem);

% Display some statistics.

figure;

semilogy([info.iter], [info.gradnorm], '.-');

xlabel('Iteration number');

ylabel('Norm of the gradient of f');

function [f, store] = mycost(x, store)

if ~isfield(store, 'Ax')

store.Ax = A*x; % The store memory is associated to a specific x

end

Ax = store.Ax;

f = -x'*Ax; % No need to cache f: cost values are cached 'under the hood'

end

function [g, store] = myegrad(x, store)

% This could be placed in a separate function

% to avoid code duplication.

if ~isfield(store, 'Ax')

store.Ax = A*x;

end

Ax = store.Ax;

% Euclidean gradient; this is also cached 'under the hood'.

g = -2*Ax;

end

[Nicolas Boumal]

Hello,

A typical approach is as follows:

Define your cost function as:  function [f, store] = mycost(x, store, A)

And add it to your problem structure like this:  problem.cost  = @(x, store) mycost(x, store, A);

Then proceed similarly for all the other functions.

I hope this helps, let us know if it's unclear.

Best,

Nicolas

[Poo Uoo]

thanks very much for your help.