# encoding=utf8
"""Implementation of various utility functions."""
import logging
from numpy import (
ndarray,
asarray,
full,
empty,
where,
random as rand,
ceil,
amin,
amax
)
logging.basicConfig()
logger = logging.getLogger("NiaPy.util.utility")
logger.setLevel("INFO")
__all__ = [
"limit_repair",
"limitInversRepair",
"objects2array",
"wangRepair",
"randRepair",
"fullArray",
"reflectRepair"
]
[docs]def limit_repair(x, Lower, Upper, **kwargs):
r"""Repair solution and put the solution in the random position inside of the bounds of problem.
Arguments:
x (numpy.ndarray): Solution to check and repair if needed.
Lower (numpy.ndarray): Lower bounds of search space.
Upper (numpy.ndarray): Upper bounds of search space.
kwargs (Dict[str, Any]): Additional arguments.
Returns:
numpy.ndarray: Solution in search space.
"""
# TODO: Add one-liner np.clip approach
ir = where(x < Lower)
x[ir] = Lower[ir]
ir = where(x > Upper)
x[ir] = Upper[ir]
return x
[docs]def limitInversRepair(x, Lower, Upper, **kwargs):
r"""Repair solution and put the solution in the random position inside of the bounds of problem.
Arguments:
x (numpy.ndarray): Solution to check and repair if needed.
Lower (numpy.ndarray): Lower bounds of search space.
Upper (numpy.ndarray): Upper bounds of search space.
kwargs (Dict[str, Any]): Additional arguments.
Returns:
numpy.ndarray: Solution in search space.
"""
ir = where(x < Lower)
x[ir] = Upper[ir]
ir = where(x > Upper)
x[ir] = Lower[ir]
return x
[docs]def wangRepair(x, Lower, Upper, **kwargs):
r"""Repair solution and put the solution in the random position inside of the bounds of problem.
Arguments:
x (numpy.ndarray): Solution to check and repair if needed.
Lower (numpy.ndarray): Lower bounds of search space.
Upper (numpy.ndarray): Upper bounds of search space.
kwargs (Dict[str, Any]): Additional arguments.
Returns:
numpy.ndarray: Solution in search space.
"""
ir = where(x < Lower)
x[ir] = amin([Upper[ir], 2 * Lower[ir] - x[ir]], axis=0)
ir = where(x > Upper)
x[ir] = amax([Lower[ir], 2 * Upper[ir] - x[ir]], axis=0)
return x
[docs]def randRepair(x, Lower, Upper, rnd=rand, **kwargs):
r"""Repair solution and put the solution in the random position inside of the bounds of problem.
Arguments:
x (numpy.ndarray): Solution to check and repair if needed.
Lower (numpy.ndarray): Lower bounds of search space.
Upper (numpy.ndarray): Upper bounds of search space.
rnd (mtrand.RandomState): Random generator.
kwargs (Dict[str, Any]): Additional arguments.
Returns:
numpy.ndarray: Fixed solution.
"""
ir = where(x < Lower)
x[ir] = rnd.uniform(Lower[ir], Upper[ir])
ir = where(x > Upper)
x[ir] = rnd.uniform(Lower[ir], Upper[ir])
return x
[docs]def reflectRepair(x, Lower, Upper, **kwargs):
r"""Repair solution and put the solution in search space with reflection of how much the solution violates a bound.
Args:
x (numpy.ndarray): Solution to be fixed.
Lower (numpy.ndarray): Lower bounds of search space.
Upper (numpy.ndarray): Upper bounds of search space.
kwargs (Dict[str, Any]): Additional arguments.
Returns:
numpy.ndarray: Fix solution.
"""
ir = where(x > Upper)
x[ir] = Lower[ir] + x[ir] % (Upper[ir] - Lower[ir])
ir = where(x < Lower)
x[ir] = Lower[ir] + x[ir] % (Upper[ir] - Lower[ir])
return x
[docs]def fullArray(a, D):
r"""Fill or create array of length D, from value or value form a.
Arguments:
a (Union[int, float, numpy.ndarray], Iterable[Any]): Input values for fill.
D (int): Length of new array.
Returns:
numpy.ndarray: Array filled with passed values or value.
"""
A = []
if isinstance(a, (int, float)):
A = full(D, a)
elif isinstance(a, (ndarray, list, tuple)):
if len(a) == D:
A = a if isinstance(a, ndarray) else asarray(a)
elif len(a) > D:
A = a[:D] if isinstance(a, ndarray) else asarray(a[:D])
else:
for i in range(int(ceil(float(D) / len(a)))):
A.extend(a[:D if (D - i * len(a)) >= len(a) else D - i * len(a)])
A = asarray(A)
return A
[docs]def objects2array(objs):
r"""Convert `Iterable` array or list to `NumPy` array.
Args:
objs (Iterable[Any]): Array or list to convert.
Returns:
numpy.ndarray: Array of objects.
"""
a = empty(len(objs), dtype=object)
for i, e in enumerate(objs):
a[i] = e
return a
