H2. Input Deck

All settings for the optimizer are defined in the input deck file. The syntax of this input deck file is defined by the Input Deck Programming Language Section 2.2.1. A syntax table is shown in Figure H.2.

Figure H.2: Possible Input Deck Programming Language settings of the donopt program.

parameters { <string> { nom = <float> ; 
                        min = <float> ; max = <float> ; }
           [ <string> { nom = <float> ; 
                        min = <float> ; max = <float> ; } ] 
           }
[ equalityConstraints { <string> {}
                      [ <string> {} ] } ]
[ inequalityConstraints { <string> {}
                        [ <string> {} ] } ]

[ target   { [ trans = "<string>" ; ] } ]
optimizer  { [ debugfile = "<string>" ; ]
             [ statfile = "<string>" ; ]
             [ del0 = <float> ; ]
             [ tau0 = <float> ; ]
             [ tau = <float> ; ]
             [ nreset = <float> ; ]
             [ epsdif = <float> ; ]
             [ gradient = <float> ; ]
             [ iotype = "<string>" ; ]
             [ chkptfile = "<string>" ; ]
           }

In the optimizer section of the input deck several parameters of the optimizer can be set. The meaning of these values is explained in Table H.1.

Table H.1: Description of the parameters in the optimizer section.

Parameter	Description
del0	In the initial phase of minimization an inequality constraint is considered binding if the scaled value of the constrained is smaller than del0. Good values are between 1.0 and 0.001. If del0 is equal to zero, it is set to the value of tau0. If del0 is chosen too small, then identification of the correct set of binding constraints may be delayed.
tau0	This parameter gives a (universal) bound describing how much the unscaled penalty-term (masimum of the constraint violation) may deviate from zero. donlp2 assumes that within the region described by tau0 all functions may be evaluated safely. The initial guess however may violate these requirements. In that case an initial feasibility improvement phase is run by donlp2 until a point is found, that fits them. tau0 may be chosen as large as a user may want. A small tau0 diminishes the efficiency of donlp2, because the iterates then will follow the boundary of the feasible set closely. Bound constraints always remain satisfied.
nreset	If there are more then nreset steps using small stepsizes and therefore small corrections, a restart of the accumulated quasi-Newton-update is tried. nreset is internally bounded by 4 from below. Good values are between n and $3 \cdot n$.
epsdif	Discretizationstepsize (if used at all) using finite difference formulas to obtain the gradients. donlp2 assumes relative precision of epsdif in the gradients.
gradient	Variable used in determining a suitable step length for the forward-difference approximation. The step length of a control parameter is determined by the minimum and maximum values of the parameter and the value of the relative step size variable gradient.
chkptfile	With this variable a file name is assigned where an image of the running program is dumped to if this is supported on this operating system (see Section 4.5.2). Omitting this variable disables this functionality.