Building Test Problems

In this section, we describe the interface for finding and compiling CUTEst problems.

Finding Problems

CUTEst has a scheme for classifying problems (see here). Based on these properties, we can search for test problems using the find_problems() function. We can check the properties of a specific problem with problem_properties().

# Example: using problem classification system
import pycutest

# Find unconstrained, variable-dimension problems
probs = pycutest.find_problems(constraints='unconstrained', userN=True)
print(sorted(probs)[:5])

# Properties of problem ROSENBR
print(pycutest.problem_properties('ROSENBR'))

Which produces the output:

# List of unconstrained, variable-dimension problems
['ARGLINA', 'ARGLINB', 'ARGLINC', 'ARGTRIGLS', 'ARWHEAD']
# Properties of problem 'ROSENBR'
{'objective': 'sum of squares', 'constraints': 'unconstrained', 'regular': True, 'degree': 2, 'origin': 'academic', 'internal': False, 'n': 2, 'm': 0}

Full documentation for these functions is given below.

Building Problems

Many CUTEst problems have optional input parameters. The print_available_sif_params() function prints a list of valid parameters for a given problem. Then, we can build a problem, including optional parameters, with import_problem(). This returns an instance of the CUTEstProblem class (see next section).

# Example: building problem
import pycutest

# Print parameters for problem ARGLALE
pycutest.print_available_sif_params('ARGLALE')

# Build this problem with N=100, M=200
problem = pycutest.import_problem('ARGLALE', sifParams={'N':100, 'M':200})
print(problem)

The available parameters for this problem, as shown by PyCUTEst, are:

Parameters available for problem ARGLALE:
N = 10 (int)
N = 50 (int)
N = 100 (int)
N = 200 (int) [default]
M = 20 (int, .ge. N)
M = 100 (int, .ge. N)
M = 200 (int, .ge. N)
M = 400 (int, .ge. N) [default]
End of parameters for problem ARGLALE
# Built problem
CUTEst problem ARGLALE (params {'M': 200, 'N': 100}) with 100 variables and 200 constraints

This means that this problem has two integer parameters N and M (default 200 and 400 respectively), where N cannot be smaller than M.

Full documentation for these functions is given below.

Cache Management

PyCUTEst works by compiling each problem in its own folder inside its cache (given by the PYCUTEST_CACHE environment variable if specified, or the current working directory if not). A problem can be cleared from the cache using clear_cache(), and a list of all problems currently installed can be displayed with all_cached_problems(). Documentation for these functions is given below.

Full function documentation

Please click on a pycutest function below for full documentation:

`find_problems`([objective, constraints, ...])	Returns the problem names of problems that match the given requirements.
`problem_properties`(problemName)	Returns problem properties (based on the CUTEst problem classification string).
`print_available_sif_params`(problemName)	Call sifdecode on given problem to print out available parameters This function is OS dependent.
`import_problem`(problemName[, destination, ...])	Prepares a problem interface module, imports and initializes it.
`clear_cache`(problemName[, sifParams])	Deletes a saved problem.
`all_cached_problems`()	Return a list of all cached problems.