Valid Ipopt options for CppAD

[type-null]

When tinkering with Ipopt options, I tried to adjust "acceptable_obj_change_tol" to terminate the program when the objective didn't improve much.

`acceptable_obj_change_tol`: "Acceptance" stopping criterion based on objective function change.

If the relative change of the objective function (scaled by Max(1,|f(x)|)) is less than this value, this part of the acceptable tolerance termination is satisfied; see also acceptable_tol. This is useful for the quasi-Newton option, which has trouble to bring down the dual infeasibility. The valid range for this real option is 0 ≤ acceptable_obj_change_tol and its default value is 10+20.

However, no matter set to 1e5000 or 1e-5000, withacceptable_iter set to 1, the iteration number of my program didn't change.

Then, I tried to append a nonsense option and it didn't give error message.

I wonder if acceptable_obj_change_tol and possibly other options are supported by CppAD or not. If so, how should I use this threshold correctly to stop the program earlier.

Many thanks!

[bradbell]

Are you using the interface described under options on ?
https://cppad.readthedocs.io/en/latest/ipopt_solve.html#options

[type-null]

Yep, here is how I used options in my code:

// object that computes objective and constraints
FG_eval fg_eval(m);

// options
string options;
// printing level: 0-12, default 5
options += "Integer print_level  5\n";
//linear solver choosing: m27, m86, m77, m97
options += "String linear_solver   ma97\n";
//derivative testing: none, first-order, second-order, only-second-order
options += "String derivative_test   none\n";

options += "Numeric acceptable_constr_viol_tol   1e-7\n";

options += "Numeric acceptable_obj_change_tol       1e50000\n";

options += "Numeric acceptable_iter       1\n";

options += "Numeric hahaha     1\n";


CppAD::ipopt::solve_result<Dvector> newSolution; // solution
CppAD::ipopt::solve<Dvector, FG_eval>(options, x0, xl, xu, gl, gu, fg_eval, newSolution); // solve the problem

[bradbell]

1. Try adding: options += "Integer print_level 5\n";
2. What is your setting for tol (the convergence tolerance) ?

[type-null]

1. That's my first options as in my last reply
2. I set it to options += "Numeric tol 1e-6\n";

How does acceptable_obj_change_tol work? Since its default value is 10^20, every iteration should have satisfied this criterion. I also set acceptable_iter to 1 so why not my program stop at 2 iterations but continues as it likes?

[bradbell]

I have not used accceptable_obj_change_tol or acceptable_iter.
I have been using tol and max_iter instead; see the heading Termination Criteria on
https://coin-or.github.io/Ipopt/OPTIONS.html#OPT_Termination

What does your optimizer trace look like when you set the print_level to 5 ?

[type-null]

Like this:

# of nonzero in Jacobin, hessian,...

iterations

exit message

[bradbell]

It should look like this:
his is Ipopt version 3.13.4, running with linear solver mumps.
NOTE: Other linear solvers might be more efficient (see Ipopt documentation).

Number of nonzeros in equality constraint Jacobian...: 0
Number of nonzeros in inequality constraint Jacobian.: 0
Number of nonzeros in Lagrangian Hessian.............: 0

Total number of variables............................: 121
variables with only lower bounds: 0
variables with lower and upper bounds: 121
variables with only upper bounds: 0
Total number of equality constraints.................: 0
Total number of inequality constraints...............: 0
inequality constraints with only lower bounds: 0
inequality constraints with lower and upper bounds: 0
inequality constraints with only upper bounds: 0

iter objective inf_pr inf_du lg(mu) ||d|| lg(rg) alpha_du alpha_pr ls
0 1.3338628e-01 0.00e+00 1.00e+00 0.0 0.00e+00 - 0.00e+00 0.00e+00 0
1 1.6012478e+00 0.00e+00 4.77e+00 -1.1 7.45e-02 - 9.98e-01 1.00e+00f 1
2 1.6503250e+00 0.00e+00 1.65e+00 -1.2 2.88e-01 - 1.00e+00 3.44e-01f 1
3 2.0201332e+00 0.00e+00 4.54e+00 -1.2 2.17e-01 - 1.00e+00 1.00e+00f 1
...

I am interested in the inf_pr and inf_du values.

[bradbell]

perhaps it would be best if you e-mail me an entire trace at [email protected]