
The SifDec package supplies a decoder which turns
SIFencoded
test problems into a collection of
Fortran 77 routines. Among the SifDec test problems are the CUTE test
problems. These problems span a wide spectrum of difficulty, ranging from
smallscale differentiable unconstrained minimization to largescale
equality and inequalityconstrained dense and sparse problems, systems of
nonlinear equations, network problems etc.
Since the SIF is a superset of the MPS modeling language, access to
suites of linear programming problems, such as the
Netlib LP
problems is possible.
Among other things, the set covers the Argonne Test Set, the
Hock and Schittkowski collection, the Dembo network
problems, the Gould QPs, etc.
The SifDec
test set is available online.
The extension of the SIF to more naturally support quadratic terms in
the involved functions resulted in our including the
Maros and
Meszaros convex quadratic programming problems set in the main
SifDec set [updated].
The default size of the SifDec problems have been raised in an
attempt to reflect the increasing size of modern problems. The
problems are now parameterized and no longer need manual
intervention. The SIF decoder now supports a commandline option
allowing to change parameters, such as the problem size, in a
SIF file.

