Skip to content

Commit

Permalink
LB-IC integer handling
Browse files Browse the repository at this point in the history
  • Loading branch information
@nikohansen
nikohansen committed Sep 3, 2024
1 parent 77e7159 commit 659fa0c
Show file tree
Hide file tree
Showing 4 changed files with 639 additions and 94 deletions.
98 changes: 42 additions & 56 deletions cma/evolution_strategy.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -205,6 +205,7 @@
from . import constraints_handler as _constraints_handler
from cma import fitness_models as _fitness_models
from .constraints_handler import BoundNone, BoundPenalty, BoundTransform, AugmentedLagrangian
from .integer_centering import IntegerCentering
from .logger import CMADataLogger # , disp, plot
from .utilities.utils import BlancClass as _BlancClass
from .utilities.utils import rglen #, global_verbosity
Expand All @@ -231,6 +232,10 @@ def _callable_to_list(c):
return [c]
return c

def _pass(*args, **kwargs):
"""a callable that does nothing and return args[0] in case"""
return args[0] if args else None

# use_archives uses collections
use_archives = sys.version_info[0] >= 3 or sys.version_info[1] >= 6
# use_archives = False # on False some unit tests fail
Expand Down Expand Up @@ -913,6 +918,7 @@ def identity(x):
# self.fmean = np.nan # TODO name should change? prints nan in output files (OK with matlab&octave)
# self.fmean_noise_free = 0. # for output only

opts.amend_integer_options(N, inopts)
self.sp = options_parameters.CMAParameters(N, opts, verbose=opts['verbose'] > 0)
self.sp0 = self.sp # looks useless, as it is not a copy

Expand Down Expand Up @@ -974,7 +980,14 @@ def eval_vector(in_, opts, N, default_value=1.0):
opts['minstd'] = eval_vector(opts['minstd'], opts, N, 0)
opts['maxstd'] = eval_vector(opts['maxstd'], opts, N, np.inf)

if 1 < 3 and len(opts['integer_variables']):
# iiinteger handling currently as LB-IC, see cma.integer_centering:
if len(opts['integer_variables']):
opts.set_integer_min_std(N, self.sp.weights.mueff)
self.integer_centering = IntegerCentering(self) # read opts['integer_variables'] and use boundary handler
else:
self.integer_centering = _pass # do nothing by default

if 11 < 3 and len(opts['integer_variables']):
try:
from . import integer
s = utils.format_message(
Expand All @@ -984,46 +997,6 @@ def eval_vector(in_, opts, N, default_value=1.0):
warnings.warn(s, category=DeprecationWarning) # TODO: doesn't show up
except ImportError:
pass
opts['integer_variables'] = list(opts['integer_variables'])

# iiinteger handling, currently very basic:
# CAVEAT: integer indices may give unexpected results if fixed_variables is used
if len(opts['integer_variables']) and opts['fixed_variables']:
# transform integer indices to genotype
popped = [] # just for the record
for i in reversed(range(self.N)):
if i in opts['fixed_variables']:
opts['integer_variables'].remove(i)
if 1 < 3: # just for catching errors
popped.append(i)
if i in opts['integer_variables']:
raise ValueError("index {0} appeared more than once in `'integer_variables'` option".format(i))
# reduce integer variable indices > i by one
for j, idx in enumerate(opts['integer_variables']):
if idx > i:
opts['integer_variables'][j] -= 1
if opts['verbose'] >= 0:
warnings.warn("Handling integer variables when some variables are fixed."
"\n This code is poorly tested and fails for negative indices."
"\n Variables {0} are fixed integer variables and discarded"
" for integer handling."
.format(popped))
# 1) prepare minstd to be a vector
if (len(opts['integer_variables']) and
np.isscalar(opts['minstd'])):
opts['minstd'] = opts['minstd'] * np.ones(N)
# 2) set minstd to 1 / (2 Nint + 1),
# the setting 2 / (2 Nint + 1) already prevents convergence
for i in opts['integer_variables']:
if -N <= i < N: # when i < 0, the index computes to N + i
if opts['minstd'][i] == 0: # negative values prevent setting minstd
# opts['minstd'][i] = 1 / (2 * len(opts['integer_variables']) + 1)
opts['minstd'][i] = min((0.2, self.sp.weights.mueff / N))
else:
utils.print_warning(
"""dropping integer index %d as it is not in range of dimension %d""" %
(i, N))
opts['integer_variables'].pop(opts['integer_variables'].index(i))

# initialization of state variables
self.countiter = 0
Expand Down Expand Up @@ -2005,22 +1978,25 @@ def get_selective_mirrors(self, number=None):

def limit_integer_relative_deltas(self, dX, threshold=None,
recombination_weight_condition=None):
"""limit absolute values of int-coordinates in vector list `dX`
"""versatile: limit absolute values of int-coordinates in vector list `dX`
relative to the current sample standard deviations and by default
only when the respective recombination weight is negative.
This function is currently not in effect (called with threshold=inf)
and not guarantied to stay as is.
``dX == pop_sorted - mold`` where ``pop_sorted`` is a genotype.
``threshold=2.3`` by default.
A 2.3-sigma threshold affects 2 x 1.1% of the unmodified
(nonsorted) normal samples.
"""
if threshold is None: # TODO: how interpret negative thresholds?
threshold = 2.3
if not self.opts['integer_variables'] or not np.isfinite(threshold):
return dX
if threshold is None: # TODO: how interpret negative thresholds?
threshold = 2.3
if recombination_weight_condition is None:
def recombination_weight_condition(w):
return w < 0
Expand Down Expand Up @@ -2300,6 +2276,8 @@ def tell(self, solutions, function_values, check_points=None,
pop = np.asarray([xp[0]] + list(pop))

self.pop_sorted = pop
self.integer_centering(pop[:sp.weights.mu], self.mean)

# compute new mean
self.mean = np.dot(sp.weights.positive_weights, pop[:sp.weights.mu])
if sp.cmean != 1:
Expand Down Expand Up @@ -2390,8 +2368,10 @@ def tell(self, solutions, function_values, check_points=None,
pass # without CSA we may not need the mean_shift

# covariance matrix adaptation/udpate
pop_zero = self.limit_integer_relative_deltas(
pop - mold, self.opts['CMA_active_limit_int_std'])
pop_zero = self.limit_integer_relative_deltas( # does by default nothing
pop - mold,
options_parameters.integer_active_limit_std,
options_parameters.integer_active_limit_recombination_weight_condition)
if c1a + cmu > 0:
# TODO: make sure cc is 1 / N**0.5 rather than 1 / N
# TODO: simplify code: split the c1 and cmu update and call self.sm.update twice
Expand Down Expand Up @@ -2429,16 +2409,22 @@ def tell(self, solutions, function_values, check_points=None,
sampler_weights[i + 1] *= self.opts['CMA_active_injected']
if sampler_weights_dd[i + 1] < 0:
sampler_weights_dd[i + 1] *= self.opts['CMA_active_injected']
for s in list(self._injected_solutions_archive):
if self._injected_solutions_archive[s]['iteration'] < self.countiter - 2:
warnings.warn("""orphanated injected solution %s
This could be a bug in the calling order/logics or due to
a too small popsize used in `ask()` or when only using
`ask(1)` repeatedly. Please check carefully.
In case this is desired, the warning can be surpressed with
``warnings.simplefilter("ignore", cma.evolution_strategy.InjectionWarning)``
""" % str(self._injected_solutions_archive.pop(s)),
InjectionWarning)
for k, s in list(self._injected_solutions_archive.items()):
if s['iteration'] < self.countiter - 2:
# warn unless TPA injections were messed up by integer centering
if (not isinstance(self.adapt_sigma, CMAAdaptSigmaTPA)
# self.integer_centering and
# self.integer_centering is not _pass and
or not isinstance(self.integer_centering, IntegerCentering)
or s['index'] > 1):
warnings.warn("""orphanated injected solution %s
This could be a bug in the calling order/logics or due to
a too small popsize used in `ask()` or when only using
`ask(1)` repeatedly. Please check carefully.
In case this is desired, the warning can be surpressed with
``warnings.simplefilter("ignore", cma.evolution_strategy.InjectionWarning)``
""" % str(s), InjectionWarning)
self._injected_solutions_archive.pop(k)
assert len(sampler_weights) == len(pop_zero) + 1
if flg_diagonal:
self.sigma_vec.update(
Expand Down
Loading

0 comments on commit 659fa0c

Please sign in to comment.