ObjectiveMerit

class kona.algorithms.util.merit.ObjectiveMerit(primal_factory, state_factory, optns={}, out_file=<open file '<stdout>', mode 'w'>)

Bases: kona.algorithms.util.merit.MeritFunction

Merit function for line searches applied to the raw objective.

Other, more complicated merit functions can be derived from this.

Variables:

• last_func_alpha (float) – Last value of alpha for which objective value is evaluated.
• last_grad_alpha (float) – Last value of alpha for which objective grad is evaluated.
• func_val (float) – Value of the objective at last_func_alpha.
• p_dot_grad (float) – Value of \(\langle p, \nabla f \rangle\) at last_grad_alpha.
• x_start (DesignVector) – Initial position of the primal variables, where \(\alpha = 0\).
• x_trial (DesignVector) – Trial position of the primal variables at a new alpha.
• u_trial (StateVector) – Trial position of the state variables at a new alpha.
• search_dir (DesignVector) – The search direction vector.
• state_work (StateVector) – Work vector for state operations.
• adjoint_work (StateVector) – Work vector for adjoint operations.
• design_work (DesignVector) – Work vector for primal operations.

eval_func(alpha)

eval_grad(alpha)

reset(search_dir, x_start, u_start, p_dot_grad)