npt is missing from the C interface

[zaikunzhang]

Hi @jschueller Julien,

I noted just now that npt was not an input of the C interface. This is an important parameter, and it should be exposed to the user. The recommended value is 2n+1, which should be documented clearly and illustrated in the examples. Powell did mention that other values may provide excellent results in some situations.

npt exists only in NEWUOA, BOBYQA, and LINCOA (COBYLA essentially fixes npt=n+1, and UOBYQA fixes npt=(n+1)(n+2)/2, and hence they are not inputs):

prima/fortran/newuoa/newuoa.f90

Lines 100 to 103 in 0afff46

	! NPT
	! Input, INTEGER(IK) scalar, default: 2N + 1.
	! NPT is the number of interpolation conditions for each trust region model. Its value must be in
	! the interval [N+2, (N+1)(N+2)/2].

prima/fortran/bobyqa/bobyqa.f90

Lines 114 to 122 in e280f51

	! NPT
	! Input, INTEGER(IK) scalar, default: 2N + 1.
	! NPT is the number of interpolation conditions for each trust region model. Its value must be in
	! the interval [N+2, (N+1)(N+2)/2]. Powell commented that "the value NPT = 2*N+1 being recommended
	! for a start ... much larger values tend to be inefficient, because the amount of routine work of
	! each iteration is of magnitude NPT**2, and because the achievement of adequate accuracy in some
	! matrix calculations becomes more difficult. Some excellent numerical results have been found in
	! the case NPT=N+6 even with more than 100 variables." And "choices that exceed 2*N+1 are not
	! recommended" by Powell.

prima/fortran/lincoa/lincoa.f90

Lines 149 to 155 in 0afff46

	! NPT
	! Input, INTEGER(IK) scalar, default: 2N + 1.
	! NPT is the number of interpolation conditions for each trust region model. Its value must be in
	! the interval [N+2, (N+1)(N+2)/2]. Typical choices of Powell were NPT=N+6 and NPT=2*N+1. Powell
	! commented in his code that "larger values tend to be highly inefficient when the number of
	! variables is substantial, due to the amount of work and extra difficulty of adjusting more
	! points."

See also discussions at emmt/OptimPack#2 (comment) .

Thanks.

Zaikun

[zaikunzhang]

closed by #68. Thanks.