MODULOPT
Optimization Routines
This page allows you to select an optimization routine from the MODULOPT Library that is suitable for solving
an optimization problem. See the list below for a brief description of
each routine.
The page describing the routine mentions the persons to contact.
Further information
Optimization routines
A routine marked with (*), called a prototype version,
is very likely to
be available only if a collaboration with the authors is accepted, so
that the authors can control the behavior of their routine.
The possible communication protocols are direct communication
(DC) or reverse communication (RC).
In DC, the optimization routine directly
calls the simulator for having information on the problem (value of
the functions and their derivatives). In RC,
the optimization routine returns to the calling routine to have
information on the problem; the calling routine calls the simulator to
get this information and loops back to call the optimization
routine and pursue the optimization.
Click on the code name to access its page and to have more information,
including the description on how to get the software.
Unconstrained optimization
Codes for solving optimization problems without constraints on the
variables.
Name
[nickname]
of the code |
Type of problem |
Algorithmic features |
Language |
Communication
Protocol |
| N1CV2 |
- convex nondifferentiable
- subgradients available
|
|
Fortran 77 |
DC |
| N1CG1 |
- large-scale convex quadratic
- Hessian-vector products available
|
- preconditioned Fletcher-Reeves conjugate gradient
- generates a (limited memory) BFGS preconditioner
|
Fortran 77 |
DC |
| M1QN3
|
- large-scale nonlinear
- first derivatives available
|
- line-search
- limited memory BFGS updates
|
Fortran 77 |
DC - RC |
(*) Prototype version
Bound constrained optimization
Codes for solving optimization problems with simple bounds on the variables
Name
[nickname]
of the code |
Type of problem |
Algorithmic features |
Language |
Communication
Protocol |
N2QP1 (*)
[QPB]
|
- medium-scale
- strictly convex quadratic
- lower bounds only
|
- Active set technique
- Gradient projection
- Cholesky factorizations
| Fortran 77 |
DC |
| M2FC1 |
- nondifferentiable
- subgradients available
|
|
Fortran 77 |
DC |
M2QN1
N2QN1
|
- medium-scale
- first derivatives available
|
- Active set technique
- BFGS updates
| Fortran 77 |
DC |
Nonlinearly constrained optimization
Codes for solving optimization problems with nonlinear constraints
on the variables
Name
[nickname]
of the code |
Type of problem |
Algorithmic features |
Language |
Communication
Protocol |
| SQPlab |
- equality and inequality constrained
- small-scale
- first/second derivatives available
|
| Matlab |
DC |