Bilevel Optimization Test Problems

Test Function Suite for Bilevel Optimization implemented in C/C++.

Wrappers available in Julia and MATLAB.

Benchmarks

Single-objective Bilevel Optimization

PMM: Five test function with pseudo-feasible solutions.
SMD: Twelve test functions, constrained and unconstrained problems.
TP: Ten test function from classical literature.

Build Project

If you are using a Linux-based operative system, you can build using the make command.

make

That command will generate the shared libraries corresponding to each test suite: pmm.so, smd.so, tp.so . That can be useful for users working on Python, R, Julia, MATLAB, etc.

Also you can run the following command (on linux) to build a shared library including all functions (make also create this library).

gcc blb18_op.c -lm -O2 -march=native  -o blb18_op.so

API References

See test.c for examples.

PMM

/*
 * D_ul: upper level dimension
 * D_ll: lower level dimension
 * x: upper level decision vector
 * y: lower level decision vector
 * fnum: function number, i.e., PMM<fnum>
 * Save in F, G the function and constraints values.
*/
void PMM_leader(int D_ul, int D_ll, double *x, double *y, double *F, double *G, int fnum)

/*
 * D_ul: upper level dimension
 * D_ll: lower level dimension
 * x: upper level decision vector
 * y: lower level decision vector
 * fnum: function number, i.e., PMM<fnum>
 * Save in f, g the function and constraints values.
*/
void PMM_follower(int D_ul, int D_ll, double *x, double *y, double *f, double *g, int fnum)

/*
 * D_ul: upper level dimension
 * D_ll: lower level dimension
 * x: upper level decision vector
 * fnum: function number, i.e., PMM<fnum>
 * Save in y the lower level optimal solution, i.e., y in argmin f(x, y)
*/
void PMM_Psi(int D_ul, int D_ll, int k, double *x, double *y, int fnum){

Contributing

Please, be free to send me your PR, issue or any comment about this repo.

jmejia8 / bilevel-benchmark