def create_training_set ( parameters, minvals, maxvals, n_train=200 ):
"""Creates a traning set for a set of parameters specified by
``parameters`` (not actually used, but useful for debugging
maybe). Parameters are assumed to be uniformly distributed
between ``minvals`` and ``maxvals``. ``n_train`` input parameter
sets will be produced, and returned with the actual distributions
list. The latter is useful to create validation sets.
Parameters
-------------
parameters: list
A list of parameter names
minvals: list
The minimum value of the parameters. Same order as ``parameters``
maxvals: list
The maximum value of the parameters. Same order as ``parameters``
n_train: int
How many training points to produce
Returns
---------
The training set and a distributions object that can be used by
``create_validation_set``
"""
distributions = []
for i,p in enumerate(parameters):
distributions.append ( ss.uniform ( loc=minvals[i], \
scale=(maxvals[i]-minvals[i] ) ) )
samples = lhd ( dist=distributions, size=n_train )
return samples, distributions
def create_training_set(parameters, minvals, maxvals, n_train=200):
"""Creates a traning set for a set of parameters specified by
``parameters`` (not actually used, but useful for debugging
maybe). Parameters are assumed to be uniformly distributed
between ``minvals`` and ``maxvals``. ``n_train`` input parameter
sets will be produced, and returned with the actual distributions
list. The latter is useful to create validation sets.
Parameters
-------------
parameters: list
A list of parameter names
minvals: list
The minimum value of the parameters. Same order as ``parameters``
maxvals: list
The maximum value of the parameters. Same order as ``parameters``
n_train: int
How many training points to produce
Returns
---------
The training set and a distributions object that can be used by
``create_validation_set``
"""
distributions = []
for i, p in enumerate(parameters):
distributions.append ( ss.uniform ( loc=minvals[i], \
scale=(maxvals[i]-minvals[i] ) ) )
samples = lhd(dist=distributions, size=n_train)
return samples, distributions
def __init__(self, rv, tag=None):
assert hasattr(rv, 'dist'), 'Input must be a distribution from ' + \
'the scipy.stats module.'
self.rv = rv
# generate the latin-hypercube points
self._mcpts = lhd(dist=self.rv, size=npts).flatten()
self.tag = tag
def __init__(self, rv, tag=None):
assert hasattr(rv, 'dist'), 'Input must be a distribution from ' + \
'the scipy.stats module.'
self.rv = rv
# generate the latin-hypercube points
self._mcpts = lhd(dist=self.rv, size=npts).flatten()
self.tag = tag
def create_training_set(parameters,
minvals,
maxvals,
fix_params=None,
n_train=200):
"""Creates a traning set for a set of parameters specified by
``parameters`` (not actually used, but useful for debugging
maybe). Parameters are assumed to be uniformly distributed
between ``minvals`` and ``maxvals``. ``n_train`` input parameter
sets will be produced, and returned with the actual distributions
list. The latter is useful to create validation sets.
It is often useful to add extra samples for regions which need to
be carefully evaluated. We do this by adding a `fix_params` parameter
which should be a dictionary indexing the parameter name, its fixed
value, and the number of additional samples that will be drawn.
Parameters
-------------
parameters: list
A list of parameter names
minvals: list
The minimum value of the parameters. Same order as ``parameters``
maxvals: list
The maximum value of the parameters. Same order as ``parameters``
fix_params: dictionary
A diciontary indexed by the parameter name. Each item will have a
tuple indicating the fixed value of the parameter, and how many
extra LHS samples are required
n_train: int
How many training points to produce
Returns
---------
The training set and a distributions object that can be used by
``create_validation_set``
"""
distributions = []
for i, p in enumerate(parameters):
distributions.append ( ss.uniform ( loc=minvals[i], \
scale=(maxvals[i]-minvals[i] ) ) )
samples = lhd(dist=distributions, size=n_train)
if fix_params is not None:
# Extra samples required
for k, v in fix_params.iteritems():
# Check whether they key makes sense
if k not in parameters:
raise ValueError, "You have specified '%s', which is" %k + \
" not in the parameters list"
extras = fix_parameter_training_set(parameters, minvals, maxvals,
k, v[0], v[1])
samples = np.r_[samples, extras]
return samples, distributions
def create_training_set ( parameters, minvals, maxvals,
fix_params=None, n_train=200 ):
"""Creates a traning set for a set of parameters specified by
``parameters`` (not actually used, but useful for debugging
maybe). Parameters are assumed to be uniformly distributed
between ``minvals`` and ``maxvals``. ``n_train`` input parameter
sets will be produced, and returned with the actual distributions
list. The latter is useful to create validation sets.
It is often useful to add extra samples for regions which need to
be carefully evaluated. We do this by adding a `fix_params` parameter
which should be a dictionary indexing the parameter name, its fixed
value, and the number of additional samples that will be drawn.
Parameters
-------------
parameters: list
A list of parameter names
minvals: list
The minimum value of the parameters. Same order as ``parameters``
maxvals: list
The maximum value of the parameters. Same order as ``parameters``
fix_params: dictionary
A diciontary indexed by the parameter name. Each item will have a
tuple indicating the fixed value of the parameter, and how many
extra LHS samples are required
n_train: int
How many training points to produce
Returns
---------
The training set and a distributions object that can be used by
``create_validation_set``
"""
distributions = []
for i,p in enumerate(parameters):
distributions.append ( ss.uniform ( loc=minvals[i], \
scale=(maxvals[i]-minvals[i] ) ) )
samples = lhd ( dist=distributions, size=n_train )
if fix_params is not None:
# Extra samples required
for k,v in fix_params.iteritems():
# Check whether they key makes sense
if k not in parameters:
raise ValueError, "You have specified '%s', which is" %k + \
" not in the parameters list"
extras = fix_parameter_training_set(parameters, minvals, maxvals,
k, v[0], v[1])
samples = np.r_[samples, extras]
return samples, distributions
def fix_parameter_training_set(parameters, minvals, maxvals,
fixed_parameter, value, n_train):
"""Produces a set of extra LHS samples where one parameter
has been fixed to a single value, whereas all other parameters
take their usual boundaries etc."""
from copy import deepcopy # groan
parameters = deepcopy(parameters)
minvals = deepcopy(minvals)
maxvals = deepcopy(maxvals)
fix_param = parameters.index(fixed_parameter)
reduced_parameters = [p for p in parameters if p != fixed_parameter]
minvals.pop(fix_param)
maxvals.pop(fix_param)
dummy_param = np.ones(n_train)*value
distributions = []
for i,p in enumerate(reduced_parameters):
distributions.append ( ss.uniform ( loc=minvals[i], \
scale=(maxvals[i]-minvals[i] ) ) )
samples = lhd ( dist=distributions, size=n_train )
extra_array = np.insert(samples, fix_param, dummy_param, axis=1)
return extra_array
def fix_parameter_training_set(parameters, minvals, maxvals, fixed_parameter,
value, n_train):
"""Produces a set of extra LHS samples where one parameter
has been fixed to a single value, whereas all other parameters
take their usual boundaries etc."""
from copy import deepcopy # groan
parameters = deepcopy(parameters)
minvals = deepcopy(minvals)
maxvals = deepcopy(maxvals)
fix_param = parameters.index(fixed_parameter)
reduced_parameters = [p for p in parameters if p != fixed_parameter]
minvals.pop(fix_param)
maxvals.pop(fix_param)
dummy_param = np.ones(n_train) * value
distributions = []
for i, p in enumerate(reduced_parameters):
distributions.append ( ss.uniform ( loc=minvals[i], \
scale=(maxvals[i]-minvals[i] ) ) )
samples = lhd(dist=distributions, size=n_train)
extra_array = np.insert(samples, fix_param, dummy_param, axis=1)
return extra_array
