5.5. pyopus.problems.mwbm
— More-Wild set of test functions for DFO, data profile generation¶
Derivative-free optimization test functions by More and Wild. (PyOPUS subsystem name: MWBM)
All 22 basic test functions [mw] in this module are maps from to . All functions are unconstrained. Every function comes from the CUTEr test set and is specified by a tuple (mgh_name, n, m, s), where n is the dimension, m is the number of component functions, and s specified the initial point scaling, i.e.
Every problem can be smooth, piecewise-smooth, deterministically noisy or stochastically noisy.
The CUTEr problems used by this suite are implemented in the _mwbm module. This module does not depend on CUTEr.
This module is independent of PyOPUS, meaning that it can be taken as is and used as a module in some other package. It depends only on the cpi and the _mwbm modules.
Beside test functions this module also includes support for data profile generation from function value history.
[mw] | More, J.J., Wild S.M., Benchmarking Derivative-Free Optimization Algorithms. SIAM Journal on Optimization, vol. 20, pp.172-191, 2009. |
-
class
pyopus.problems.mwbm.
MWBM
(num, problemType=0, epsilon=0.001)¶ Test functions from the More-Wild suite.
num is the problem number (0-52).
problemType is the type of the problem (0=smooth, 1=piecewise smooth, 2=deterministically noisy, 3=stochastically noisy).
epsilon is the noise level for problemType 2 and 3.
The number of the underlying function (0-21) is stored in
probNdx
. The dimension, the number of component functions, and the initial point scaling are stored in then
,m
, ands
members.The full name of the problem is in the
name
member. It is of the form name_m_s.The initial point can be obtained from the
initial
member. The function value is returned by thef()
method.The gradient and the best known minimum of the function are not available.
-
cpi
()¶ Returns the common problem interface.
xmin, fmin, and g are not available.
The info member of the returned dictionary is itself a dictionary with the following members:
m
- m parameter of the MWBM problems
- s parameter of the MWBM problem
See the
CPI
class for more information.
-
descriptors
= [[1, 9, 45, 0], [1, 9, 45, 1], [2, 7, 35, 0], [2, 7, 35, 1], [3, 7, 35, 0], [3, 7, 35, 1], [4, 2, 2, 0], [4, 2, 2, 1], [5, 3, 3, 0], [5, 3, 3, 1], [6, 4, 4, 0], [6, 4, 4, 1], [7, 2, 2, 0], [7, 2, 2, 1], [8, 3, 15, 0], [8, 3, 15, 1], [9, 4, 11, 0], [10, 3, 16, 0], [11, 6, 31, 0], [11, 6, 31, 1], [11, 9, 31, 0], [11, 9, 31, 1], [11, 12, 31, 0], [11, 12, 31, 1], [12, 3, 10, 0], [13, 2, 10, 0], [14, 4, 20, 0], [14, 4, 20, 1], [15, 6, 6, 0], [15, 7, 7, 0], [15, 8, 8, 0], [15, 9, 9, 0], [15, 10, 10, 0], [15, 11, 11, 0], [16, 10, 10, 0], [17, 5, 33, 0], [18, 11, 65, 0], [18, 11, 65, 1], [19, 8, 8, 0], [19, 10, 12, 0], [19, 11, 14, 0], [19, 12, 16, 0], [20, 5, 5, 0], [20, 6, 6, 0], [20, 8, 8, 0], [21, 5, 5, 0], [21, 5, 5, 1], [21, 8, 8, 0], [21, 10, 10, 0], [21, 12, 12, 0], [21, 12, 12, 1], [22, 8, 8, 0], [22, 8, 8, 1]]¶ Problem descriptors (problem number, n, m, s)
-
namelist
= ['LinearFullRank', 'LinearRank1', 'LinearRank1ZCR', 'Rosenbrock', 'HelicalValley', 'PowellSingular', 'FreudensteinAndRoth', 'Bard', 'KowalikAndOsborne', 'Meyer', 'Watson', 'Box3D', 'JennrichAndSampson', 'BrownAndDennis', 'Chebyquad', 'BrownAlmostLinear', 'Osborne1', 'Osborne2', 'Bdqrtic', 'Cube', 'Mancino', 'Heart8']¶ List of problem names
-
Example file mwbm.py in folder demo/problems/
# Demo of the More-Wild problems
from pyopus.problems.mwbm import MWBM
if __name__ == '__main__':
# Every problem is described by an entry in the list of descriptors.
# We use this list to get a problem count.
for ii in range(len(MWBM.descriptors)):
# Create the default (smooth) problem
problem=MWBM(ii, problemType=0)
# Set problemType to
# 1 for piecewise-smooth problems
# 2 for deterministically noisy problems
# 3 for stochastically noisy problems
cpi=problem.cpi()
print(
"%2d: %25s n=%2d m=%2d s=%2d f0=%e" %
(
ii,
cpi['name'],
cpi['n'],
cpi['info']['m'],
cpi['info']['s'],
cpi['f'](cpi['x0'])
)
)