Open jfowkes opened 10 months ago
For now I will prioritise (all for constrained problems):
cdhj
: evaluate the dense Hessian of the (Fritz) John functioncshj
: evaluate the sparse Hessian of the (Fritz) John functionchjprod
: product of Hessian of the (Fritz) John function with a vectorcohprods
: sparse product of (objective) Hessian with a vectoras we need these for some of our research.
Might also be good to support the following existing CUTEst routines:
ufn
: evaluate objective functioncifn
: evaluate objective function or constraint functioncdhc
: evaluate the Hessian of the constraint part of the Lagrangianchcprod
: product of Hessian of the constraint part of the Lagrangian with a vectorcchprods
: products of Hessians of the constraint with a vectorcshc
: evaluate the Hessian of the constraint part of the Lagrangian (sparse)These were either not used or not in CUTEr (the precursor to CUTEst).
CUTEst 2.2 added routines for the (Fritz) John function, defined as
$$ \mathcal{J}(x,\lambda_0,\lambda) = \lambda_0 f(x) + \langle \lambda, c(x) \rangle $$
(for more details please see https://en.wikipedia.org/wiki/Fritz_John_conditions)
The following new routines have been added to CUTEst 2.2:
We should add support for these in PyCUTEst as appropriate.