"""
.. inheritance-diagram:: pyopus.evaluator.aggregate
:parts: 1
**Parformance aggregation module (PyOPUS subsystem name: AG)**
**Normalization** is the process where a performance measure is scaled in such
way that values not satifying the goal result in normalized values smaller than
0. If the performance measure value exceeds the goal the normalized value is
greater than 0. Complex values are treated as failures and failure penalty is
used for norm.
**Shaping** is the process where the normalized performance is shaped.
Usually positive values (corresponding to performance not satifying the goal)
are shaped differently than negative values (corresponding to performance
satifying the goal). Shaping results in a **shaped contribution** for
every corner.
**Corner reduction** is the process where **shaped contributions** of
individual corners are incorporated into the aggregate function. There are
several ways how to achieve this. For instance one could incorporate only the
contribution of the corner in which worst peroformance is observed, or on the
other hand one could incorporate the mean contribution of all corners.
The main data structure is the **aggregate function description** which is a list
of **component descriptions**.
Every **component description** is a dictionary with the following members:
* ``measure`` - the name of the performance meeasure on which the aggregate
function's component is based.
* ``norm`` - an object performing the normalization of the performance measure.
* ``shape`` - an object performing the shaping of normalized performance measure.
Defaults to ``Slinear2(1.0, 0.0)``.
* ``reduce`` - an object performing the corner reduction of the cotributions.
Defaults to ``Rworst()``.
The **ordering of parameters** is a list of parameter names that defines the
order in which parameter values apper in a parameter vector.
A **parameter vector** is a list or array of parameter values where the values
are ordered according to a given **ordering of input parameters**.
"""
from ..optimizer.base import Reporter, Stopper, Plugin, CostCollector, Annotator
from ..misc.debug import DbgMsgOut, DbgMsg
from .auxfunc import paramList, paramDict
from .. import PyOpusError
from numpy import concatenate, array, ndarray, where, zeros, sum, abs, array, ndarray, iscomplex
import sys
normalizationNames = [ 'Nbase', 'Nabove', 'Nbelow', 'Nbetween' ]
shaperNames = [ 'Slinear2' ]
reductionNames = [ 'Rbase', 'Rexcluded', 'Rworst', 'Rmean' ]
__all__ = normalizationNames + shaperNames + reductionNames + [
'formatParameters', 'Aggregator',
]
# Basic normalization class
[docs]class Nbase(object):
"""
Basic normalization class. Objects of this class are callable. The calling
convention of the object is ``object(value)`` where *value* is a scalar or
an array of performance measure values. When called with a scalar the
return value is a scalar. When called with an array the return value is an
array of the same shape where every component is treated as if it was a
scalar.
The return value is greater than 0 if the passed value fails to satisfy the
goal. It is less than 0 if the passed value exceeds the goal. Exceeding the
goal by *norm* means that the return value is -1.0. Failing to satify the
goal by *norm* results in a return value of 1.0. If the value passed at
call is ``None``, the return value is equal to *failure*.
If *norm* is not given the default normalization is used which is equal to
*goal* or 1.0 if *goal* is equal to 0.0. If *goal* is a vector, *norm*
must either be a vector of the same size or a scalar in which case it
applies to all components of the *goal*.
"""
def __init__(self, goal, norm=None, failure=10000.0):
self.goal=array(goal)
if norm is None:
if self.goal.size==1:
self.norm=abs(self.goal) #/10.0
if self.norm==0:
self.norm=1.0
else:
self.norm=abs(self.goal) #/10.0
self.norm[where(self.norm==0.0)]=1.0
else:
self.norm=norm
self.failure=failure
def __call__(self, value):
pass
[docs] def worst(self, values, total=False):
"""
Returns the worst performance value across all corners (the one with
the largest normalized value). The values across corners are given in
the *values* array where first array index is the corner index.
If the array has more than 1 dimension the worst value is sought along
the first dimension of the array. This means that if *value* is of
shape (n, m1, m2, ...) then the return value is of shape (m1, m2, ...).
The return value is an array of performance measure values.
If *total* is ``True`` the worst value is sought acros the whole array
and the return value is a scalar worst performance value.
"""
pass
[docs] def worstCornerIndex(self, values, corners, total=False):
"""
Returns the index corresponding to the corner where the performance
measure takes its worst value (the one with the largest normalized
value). The values across corners are given in the *values* array where
first array index is the corner index.
If the array has more than 1 dimension the worst value is sought along
the first dimension of the array. This means that if *value* is of
shape (n, m1, m2, ...) then the return value is of shape (m1, m2, ...).
The return value is an array of corner indices.
The corner indices corresponding to the first dimension of *values* are
given by *corners*.
If *total* is ``True`` the worst value is sought across the whole array
and the return value is a scalar worst corner index.
"""
pass
[docs] def report(self, name, nName=12, nGoal=12, nSigGoal=3):
"""
Formats the goal as a string of the form
*name* *normalization_symbol* *goal* where *name*
is the name of the performance measure. The *normalization_symbol*
depends on the type of normalization (derived class).
*nName* and *nGoal* specify the width of performance measure name and
goal formatting. *nSigGoal* is the number of significant digits of the
*goal* in the formatted string.
"""
pass
# Normalization for targets of the form value>goal.
[docs]class Nabove(Nbase):
"""
Performance normalization class requiring the performance to to be above
the given *goal*. See :class:`Nbase` for more information.
"""
def __init__(self, goal, norm=None, failure=10000.0):
Nbase.__init__(self, goal, norm, failure)
def __str__(self):
return "> %e" % (self.goal)
[docs] def worst(self, values, total=False):
"""
Find the worst *value*. See :meth:`Nbase.worst` method for more
information.
"""
if not total:
return values.min(0)
else:
return values.min()
[docs] def worstCornerIndex(self, values, corners, total=False):
"""
Find the worst corner index. See :meth:`Nbase.worstCornerIndex` method
for more information.
"""
if not total:
return corners[values.argmin(0)]
else:
return corners.ravel()[values.argmin()]
def __call__(self, value):
return (self.goal-value)/self.norm
[docs] def report(self, name, nName=12, nGoal=12, nSigGoal=3):
"""
Format the goal as a string. The output is a string of the form
*name* > *goal*
See :meth:`Nbase.report` method for more information.
"""
if self.goal.size!=1:
return "%*s > %-*s" % (nName, name, nGoal, 'vector')
else:
return "%*s > %-*.*e" % (nName, name, nGoal, nSigGoal, self.goal)
# Normalization for targets of the form value<goal.
[docs]class Nbelow(Nbase):
"""
Performance normalization class requiring the performance to to be below
the given *goal*. See :class:`Nbase` for more information.
"""
def __init__(self, goal, norm=None, failure=10000.0):
Nbase.__init__(self, goal, norm, failure)
def __str__(self):
return "< %e" % (self.goal)
[docs] def worst(self, values, total=False):
"""
Find the worst *value*. See :meth:`Nbase.worst` method for more
information.
"""
if not total:
return values.max(0)
else:
return values.max()
[docs] def worstCornerIndex(self, values, corners, total=False):
"""
Find the worst corner index. See :meth:`Nbase.worstCornerIndex` method
for more information.
"""
if not total:
return corners[values.argmax(0)]
else:
return corners.ravel()[values.argmax()]
def __call__(self, value):
return (value-self.goal)/self.norm
[docs] def report(self, name, nName=12, nGoal=12, nSigGoal=3):
"""
Format the goal as a string. The output is a string of the form
*name* < *goal*
See :meth:`Nbase.report` method for more information.
"""
if self.goal.size!=1:
return "%*s < %-*s" % (nName, name, nGoal, 'vector')
else:
return "%*s < %-*.*e" % (nName, name,nGoal, nSigGoal, self.goal)
# Normalization for targets of the form goal<value<goalHigh.
[docs]class Nbetween(Nbase):
"""
Performance normalization class requiring the performance to to be above
*goal* and below *goalHigh*. See :class:`Nbase` for more information.
This class is deprecated. Use two contributions instead (one with
Nbelow and one with Nabove).
"""
def __init__(self, goal, goalHigh, norm=None, failure=10000.0):
Nbase.__init__(self, goal, norm, failure)
self.goalHigh=array(goalHigh)
if (self.goal>self.goalHigh).any():
raise PyOpusError(DbgMsg("AG", "Lower bound is above upper bound."))
if norm is None:
if self.goal.size==1:
self.norm=abs(self.goalHigh-self.goal) # /10.0
if self.norm==0:
self.norm=1.0
else:
self.norm=abs(self.goalHigh-self.goal) # /10.0
self.norm[where(self.norm==0.0)]=1.0
else:
self.norm=norm
[docs] def worst(self, values, total=False):
"""
Find the worst *value*. See :meth:`Nbase.worst` method for more
information.
"""
# Distance from center
distance=abs(values-(self.goal+self.goalHigh)/2)
if values.size==1:
return values
else:
if not total:
return values[distance.argmax(0)]
else:
return values[distance.argmax()]
[docs] def worstCornerIndex(self, values, corners, total=False):
"""
Find the worst corner index. See :meth:`Nbase.worstCornerIndex` method
for more information.
"""
# Distance from center
distance=abs(values-(self.goal+self.goalHigh)/2)
if not total:
return corners[distance.argmax(0)]
else:
return corners.ravel()[distance.argmax()]
def __call__(self, value):
center=(self.goal+self.goalHigh)/2
return where(value<center, (self.goal-value)/self.norm, (value-self.goalHigh)/self.norm)
[docs] def report(self, name, nName=12, nGoal=12, nSigGoal=3):
"""
Format the goal as a string. The output is a string of the form
*name* < > (*goal* + *goalHigh*)/2
See :meth:`Nbase.report` method for more information.
"""
if self.goal.size!=1:
return "%*s < > %-*s" % (nName, name, nGoal, 'vector')
else:
return "%*s < > %-*.*e" % (nName, name, nGoal, nSigGoal, (self.goal+self.goalHigh)/2)
# Linear two-segment shaping
[docs]class Slinear2(object):
"""
Two-segment linear shaping. Normalized performances above 0 (failing
to satify the goal) are multiplied by *w*, while the ones below 0
(satisfying the goal) are multiplied by *tw*. This shaping has a
discontinuous first derivative at 0.
Objects of this class are callable. The calling comvention is
``object(value)`` where *value* is an array of normalized performance
measures.
"""
def __init__(self, w=1.0, tw=0.0):
self.w=w
self.tw=tw
def __str__(self):
return "wbad=%e wgood=%e" % (self.w, self.tw)
def __call__(self, normMeasure):
goodI=where(normMeasure<=0)
badI=where(normMeasure>0)
shapedMeasure=normMeasure.copy()
shapedMeasure[goodI]*=self.tw
shapedMeasure[badI]*=self.w
return shapedMeasure
# Basic corner reduction class
# Reducing contributions from multiple corners to one.
# The corners are the first dimension of the input array.
[docs]class Rbase(object):
"""
Basic corner reduction class. Objects of this class are callable. The
calling convention of the object is ``object(shapedMeasure)`` where
*shapedMeasure* is an array of shaped contributions. The return value
is a scalar.
"""
def __init__(self):
pass
def __call__(self, shapedMeasure):
pass
[docs] def flagSuccess(self, fulfilled):
"""
Return a string that represents a flag for marking performance measures
that satify the goal (*fulfilled* is ``True``) or fail to satisfy the
goal (*fulfilled* is ``False``).
"""
pass
[docs] def flagFailure(self):
"""
Return a string that represents the flag for marking performance
measures for which the process of evaluation failed (their value is
``None``).
"""
pass
# Excludes contribution, i.e. always returns 0
[docs]class Rexcluded(Rbase):
"""
Corner reduction class for excluding the performance measure from the
aggregate function. Objects of this class are callable and return 0.
See :class:`Rbase` for more information.
"""
def __init__(self):
Rbase.__init__(self)
def __call__(self, shapedMeasure):
return array(0.0)
def __str__(self):
return "excluded"
# Returns a characters for output.
# ' ' if fulfilled is true, '.' if fulfilled is false.
[docs] def flagSuccess(self, fulfilled):
"""
Return a string that represents a flag for marking performance
measures. A successfully satisfied goal is marked by ``' '`` while a
failure is marked by ``'.'``.
See :meth:`Rbase.flagSuccess` method for more information.
"""
if fulfilled:
return ' '
else:
return '.'
# Returns a character for failure ('x').
[docs] def flagFailure(self):
"""
Return a string that represents the flag for marking performance
measures for which the process of evaluation failed (their value is
``None``). Returns ``'x'``.
See :meth:`Rbase.flagFailure` method for more information.
"""
return 'x'
# Returns largest contribution across corners
[docs]class Rworst(Rbase):
"""
Corner reduction class for including only the worst performance measure
across all corners. Objects of this class are callable and return the
larget contribution.
*component* specifies the component of a vector to be included in the
aggregate cost function. If set to ``None`` all components are considered.
See :class:`Rbase` for more information.
"""
def __init__(self, component=None):
self.component=component
# Reduce along first axis (corners)
def __call__(self, shapedMeasure):
return shapedMeasure.max(0) if self.component is None else shapedMeasure.max(0)[self.component]
def __str__(self):
return "worst"
[docs] def flagSuccess(self, fulfilled):
"""
Return a string that represents a flag for marking performance
measures. A successfully satisfied goal is marked by ``' '`` while a
failure is marked by ``'o'``.
See :meth:`Rbase.flagSuccess` method for more information.
"""
if fulfilled:
return ' '
else:
return 'o'
[docs] def flagFailure(self):
"""
Return a string that represents the flag for marking performance
measures for which the process of evaluation failed (their value is
``None``). Returns ``'X'``.
See :meth:`Rbase.flagFailure` method for more information.
"""
return 'X'
# Returns mean contribution.
[docs]class Rmean(Rbase):
"""
Corner reduction class for including only the mean contribution across
all corners. Objects of this class are callable and return the mean of
contributions passed at call.
See :class:`Rbase` for more information.
"""
def __init__(self, component=None):
self.component=component
# Reduce along first axis (corners)
def __call__(self, shapedMeasure):
return shapedMeasure.mean(0) if self.component is None else shapedMeasure.mean(0)[self.component]
def __str__(self):
return "mean"
[docs] def flagSuccess(self, fulfilled):
"""
Return a string that represents a flag for marking performance
measures. A successfully satisfied goal is marked by ``' '`` while a
failure is marked by ``'o'``.
See :meth:`Rbase.flagSuccess` method for more information.
"""
if fulfilled:
return ' '
else:
return 'o'
[docs] def flagFailure(self):
"""
Return a string that represents the flag for marking performance
measures for which the process of evaluation failed (their value is
``None``). Returns ``'X'``.
See :meth:`Rbase.flagFailure` method for more information.
"""
return 'X'
normalizations=set([])
for name in normalizationNames:
normalizations.add(eval(name))
shapers=set([])
for name in shaperNames:
shapers.add(eval(name))
reductions=set([])
for name in reductionNames:
reductions.add(eval(name))
[docs]class Aggregator:
"""
Aggregator class. Objects of this class are callable. The calling
convention is ``object(paramVector)`` where *paramvector* is a list or an
array of input parameter values. The ordering of input parameters is given
at object construction. The return value is the value of the aggregate
function.
*perfEval* is an object of the
:class:`~pyopus.evaluator.performance.PerformanceEvaluator`
class which is used for evaluating the performance measures of the system.
*inputOrder* is the ordering of system's input parameters.
*definition* is the aggregate function description.
If *useOnlyListedCorners* is set to ``True`` the cost function will be
constructed from those corners that are listed in measure definitions.
If no corners are listed all evaluated corners are used for constructing
the cost function. When set to ``False`` all evaluated corners are used.
If *debug* is set to a value greater than 0, debug messages are generated
at the standard output.
Objects of this class store the details of the last evaluated aggregate
function value in the :attr:`results` member which is a list (one member
for every aggregate function component) of dictionaries with the following
members:
* ``worst`` - the worst value of corresponding performance mesure across
corners where the performance measure was computed. This is the return
value of the normalization object's :meth:`Nbase.worst` method when
called with with *total* set to ``True``. ``None`` if performance measure
evaluation fails in at least one corner
* ``worst_vector`` - a vector with the worst values of the performance
measure. If the performance measure is a scalar this is also a scalar.
If it is an array of shape (m1, m2, ...) then this is an array of the
same shape. This is the return value of the normalization object's
:meth:`Nbase.worst` method with *total* set to ``False``. ``None`` if
performance measure evaluation fails in at least one corner.
* ``worst_corner`` - the index of the corner in which the worst value of
performance measure occurs. If the performance measure is an array of
shape (m1, m2, ..) this is still a scalar which refers to the corner
index of the worst performance measure across all components of the
performance measure in all corners. This is the return value of the
normalization object's :meth:`Nbase.worstCornerIndex` method with
*total* set to ``True``. If the performance evaluation fails in at least
one corner this is the index of one of the corners where the failure
occurred.
* ``worst_corner_vector`` - a vector of corner indices where the worst
value of the performance measure is found. If the performance measure is
a scalar this vector has only one component. If the performance measure
is an array of shape (m1, m2, ...) this is an array with the same shape.
This is the return value of the normalization object's
:meth:`Nbase.worstCornerIndex` method with *total* set to ``False``.
If the evaluation of a performance measure fails in at least one corner
this vector holds the indices of corners in which the failure occured.
* ``contribution`` - the value of the contribution to the aggregate
function. This is always a number, even if the evaluation of some
performance measures fails (see *failure* argument to the constructor of
normalization objects - e.g. :class:`Nbase`).
* ``fulfilled`` - ``True`` if the corresponding performance measure is
successfully evaluated in all of its corresponding corners and all
resulting values satisfy the corresponding goal. ``False`` otherwise.
Corner indices refer to corners in the *cornerOrder* member which is a
list of names of corners defined in *perfEval* (see
:class:`~pyopus.evaluator.performance.PerformanceEvaluator`).
The *cornerOrder* is in fact the *cornerOrder* member of *perfEval*.
The *measure2corner* dictionary of corner lists from the *perfEval* is
used for computing the global corner indices which correspond to
corners in the *cornerOrder* member.
The :attr:`paramVector` member holds the input parameter values passed at
the last call to this object.
"""
# Constructor
def __init__(self, perfEval, definition, inputOrder=None, useOnlyListedCorners=False, debug=0):
# Performance evaluator
self.perfEval=perfEval
# List of names of all defined corners
self.cornerOrder=perfEval.cornerOrder
self.useOnlyListedCorners=useOnlyListedCorners
# Dictionary for converting corner name to global corner index
cornerName2index={}
ii=0
for corner in self.cornerOrder:
cornerName2index[corner]=ii
ii+=1
# List of measures that appear in components
measureList=set()
for comp in definition:
measureList.add(comp['measure'])
# Conversion tables from local corner index to global corner index
self.localCI2globalCI={}
for measureName in measureList:
measure=self.perfEval.measures[measureName]
if self.useOnlyListedCorners and 'corners' in measure:
corners=measure['corners']
else:
corners=self.perfEval.measure2corners[measureName]
globalIndex=[]
for corner in corners:
globalIndex.append(cornerName2index[corner])
self.localCI2globalCI[measureName]=array(globalIndex)
# Debug mode flag
self.debug=debug
# Problem definition
self.inputOrder=inputOrder
self.costDefinition=definition
# Verify definition, set defaults
ii=0
for contribution in definition:
name=contribution['measure']
if 'norm' in contribution and type(contribution['norm']) not in normalizations:
raise PyOpusError(DbgMsg("AG", "Bad normalization for contribution %d (%s)" % (ii, name)))
if 'shape' in contribution:
if type(contribution['shape']) not in shapers:
raise PyOpusError(DbgMsg("AG", "Bad shaper for contribution %d (%s)" % (ii, name)))
else:
contribution['shape']=Slinear2(1.0, 0.0)
if 'reduce' in contribution:
if type(contribution['reduce']) not in reductions:
raise PyOpusError(DbgMsg("AG", "Bad reduction for contribution %d (%s)" % (ii, name)))
else:
contribution['reduce']=Rworst()
ii+=1
# Input parameters
self.paramVector=None
# Results of the aggregate function evaluation
self.results=None
[docs] def resolveCornerIndex(self, ndx):
"""
Returns the corner name for the given corner index.
"""
return self.cornerOrder[ndx]
def __call__(self, paramVector):
if self.debug:
DbgMsgOut("AG", "Evaluation started.")
# Store parameters
self.paramVector=array(paramVector)
# Create parameter dictionary
params=paramDict(paramVector, self.inputOrder)
if self.debug:
DbgMsgOut("AG", " Evaluating measures.")
# Evaluate performance
performances, anCount = self.perfEval(params)
if self.debug:
DbgMsgOut("AG", " Processing")
# Evaluate aggregate function
results=[]
cf=0;
# Loop through all components of the aggregate function
for component in self.costDefinition:
measureName=component['measure']
# Get performance across corners
performance=performances[measureName]
# Measure info from measures dictionary of the performance evaluator
measure=self.perfEval.measures[measureName]
# If a measure has no corners defined use the list of all corner names
if self.useOnlyListedCorners and 'corners' in measure:
measureCornerList=measure['corners']
else:
measureCornerList=self.perfEval.measure2corners[measureName]
if self.debug:
DbgMsgOut("AG", " "+str(measureName))
# If measure is a vector with m elements, it behaves as m independent measurements
# The worst_vector value of a measure across corners is a vector of worst values of
# individual vector components across corners. It is a scalar for scalar measures.
# The worst value is the worst value in the the worst_vector.
# The worst_corner_vector is a vector of corner indices
# (based on the corner ordering in the cornerOrder member)
# where every index corresponds to one component of the worst_vector.
# If the worst value occurs in several corners at the same time the index of the corner
# that appears first is stored in worst_corner_vector.
# The worst_corner is the name of the corner with the lowest index and most appearances
# in the worst_corner_vector.
# Collect measure values vector across corners
failedCorners=[]
resultCorners=[]
resultVector=[]
normWorst=None
normMean=0
for index in range(0, len(measureCornerList)):
cornerName=measureCornerList[index]
value=performance[cornerName]
if value is None:
# This worked in 3.4, but does not in 3.6 (crashes CBD)
# Only scalar arrays can be used as index for a list, vector arrays with one component cannot be used
# failedCorners.append(array([index]))
failedCorners.append(index)
else:
resultCorners.append(index)
resultVector.append(array(value))
# If measure is a vector (numpy array) resultVector is a list of n-d
# arrays (they do not have to be of same size, if the size does not
# match then the last element is multiplied until full size is reached).
# Joining them means we obtain another array with n+1 dimensions (the
# first dimension is the corner index and the remaining dimensions are
# measure vector indices).
max_dim = 0
for index in range(len(resultVector)):
if max_dim < resultVector[index].size:
max_dim = resultVector[index].size
for index in range(len(resultVector)):
if max_dim > resultVector[index].size:
tmp = zeros(max_dim - resultVector[index].size) + resultVector[index][-1]
resultVector[index] = concatenate((resultVector[index], tmp))
resultVector=array(resultVector)
resultCorners=array(resultCorners)
# Total number of corners
nFailedCorners=len(failedCorners)
nGoodCorners=len(resultCorners)
nCorners=nFailedCorners+nGoodCorners
# If a measure failed in some corner, the worstValueVector is simply None
# and the worstCornerVector is the list of corners where failure occured.
# Get worst value and corner(s)
if len(failedCorners)>0:
worstValueVector=None
worstCornerVector=array(failedCorners)
worstValue=None
worstCorner=worstCornerVector[0]
else:
# Worst value vector (across corners)
worstValueVector=component['norm'].worst(resultVector)
# Worst corner vector
worstCornerVector=component['norm'].worstCornerIndex(resultVector, resultCorners)
# Worst value
worstValue=component['norm'].worst(worstValueVector, True)
# Worst corner
worstCorner=component['norm'].worstCornerIndex(worstValueVector, worstCornerVector, True)
# Warning... corners where measurements were successfull come first, followed
# by corners where measurements failed.
# Calculate normalized measure values
normMeasureFailed=zeros(nFailedCorners)
normMeasureGood=zeros(nGoodCorners)
# Add failed corners
normMeasureFailed[:]=component['norm'].failure
# Add remaining corners (the ones where measure didn't fail)
normMeasureGood=component['norm'](resultVector)
normMeasureGood=where(iscomplex(normMeasureGood), component['norm'].failure, normMeasureGood)
# Check if the measure is fulfilled (in all corners)
if len(failedCorners)<=0:
if normMeasureGood.max()<=0:
fulfilled=True
else:
fulfilled=False
else:
fulfilled=False
# Shape normalized measure values
shapedMeasureFailed=component['shape'](normMeasureFailed)
shapedMeasureGood=component['shape'](normMeasureGood)
# Reduce multiple corners to a single value
# The failed part is just added up
cfPartFailed=shapedMeasureFailed.sum()
# This is still a vector if the measure is a vector
if nGoodCorners>0:
reduced=component['reduce'](shapedMeasureGood)
if reduced.size>1:
cfPartGood=reduced.sum()
else:
cfPartGood=reduced
else:
cfPartGood=0.0
# Add up the shaped part and the failed part
cfPart=cfPartGood+cfPartFailed
# Convert corner indices from measure corner index to global corner index
convTable=self.localCI2globalCI[measureName]
worstCorner=convTable[worstCorner]
if type(worstCornerVector) is ndarray:
for ii in range(worstCornerVector.size):
worstCornerVector[ii]=convTable[worstCornerVector[ii]]
else:
worstCornerVector=convTable[worstCornerVector]
# Put it in results structure
thisResult={}
thisResult['worst']=worstValue
thisResult['worst_corner']=worstCorner
thisResult['worst_vector']=worstValueVector
thisResult['worst_corner_vector']=worstCornerVector
thisResult['contribution']=cfPart
thisResult['fulfilled']=fulfilled
results.append(thisResult)
cf+=cfPart
self.results=results
return cf
[docs] def allFulfilled(self):
"""
Returns ``True`` if the performance measures corresponding to all
aggregate function components that were evaluated with the last call
to this object were successfully evaluated and fulfill their
corresponding goals. All components are taken into account, even
those using the :class:`Rexcluded` corner reduction.
"""
for result in self.results:
if not result['fulfilled']:
return False
return True
[docs] def allBelowOrAtZero(self):
"""
Returns ``True`` if all components of the aggregate function computed
with the last call to this object are not greater than zero.
Assumes that the following holds:
The return value is ``True``, if all performance measures corresponding
to aggregate function components not using the :class:`Rexcluded`
corner reduction satisfy their goals; assuming that
* normalization produces positive values for satisfied goals and
negative values for unsatisfied goals
* normalization returns a positive value in case of a failure to
evaluate a performance measure (*failed* is greater than 0)
* aggregate function shaping is nondecreasing and is greater than zero
for positive normalized performance measures
"""
for result in self.results:
if result['contribution']>0:
return False
return True
# Return annotator plugin.
[docs] def getAnnotator(self):
"""
Returns an object of the :class:`CostAnnotator` class which can be used
as a plugin for iterative algorithms. The plugin takes care of aggregate
function details (:attr:`results` member) propagation from the machine
where the evaluation of the aggregate function takes place to the machine
where the evaluation was requested (usually the master).
"""
return AggregatorAnnotator(self)
# Return collector plugin.
[docs] def getCollector(self, chunkSize=10):
"""
Returns an object of the :class:`CostCollector` class which can be
used as a plugin for iterative algorithms. The plugin gathers input
parameter and aggregate function values across iterations of the
algorithm.
*chunkSize* is the chunk size used when allocating space for stored
values (10 means allocation takes place every 10 iterations).
"""
return CostCollector(self, chunkSize)
# Return stopper plugin that stops optimization when all requirements are satisfied.
[docs] def getStopWhenAllSatisfied(self):
"""
Returns an object of the :class:`StopWhenAllSatisfied` class which can
be used as a plugin for iterative algorithms. The plugin signals the
iterative algorithm to stop when all contributions obtained with
the last call to this :class:`Aggregator` object are smaller than
zero (when the :meth:`allBelowOrAtZero` method returns ``True``).
"""
return StopWhenAllSatisfied(self)
# Return reporter plugin.
[docs] def getReporter(self, reportParameters=True):
"""
Returns an object of the :class:`ReportCostCorners` class which can be
used as a plugin for iterative algorithms. Every time an iterative
algorithm calls this :class:`Aggregator` object the reporter is
invoked and prints the details of the aggregate function components.
"""
return AggregatorReporter(self, reportParameters)
# Default annotator for Aggregator
class AggregatorAnnotator(Annotator):
"""
A subclass of the :class:`~pyopus.optimizer.base.Annotator` iterative
algorithm plugin class. This is a callable object whose job is to
* produce an annotation (details of the aggregate function value) stored
in the *aggregator* object
* update the *aggregator* object with the given annotation
Annotation is a copy of the :attr:`results` member of *aggregator*.
Annotators are used for propagating the details of the aggregate function
from the machine where the evaluation takes place to the machine where the
evaluation was requested (usually the master).
"""
def __init__(self, aggregator):
self.ce=aggregator
def produce(self):
return self.ce.results
def consume(self, annotation):
self.ce.results=annotation
# Stopper that stops the algorithm when all requirements are satisfied
class StopWhenAllSatisfied(Stopper, AggregatorAnnotator):
"""
A subclass of the :class:`~pyopus.optimizer.base.Stopper` iterative
algorithm plugin class that stops the algorithm when the
:meth:`Aggregator.allBelowOrAtZero` method of the *Aggregator* object
returns ``True``.
This class is also an annotator that collects the results at remote
evaluation and copies them to the host where the remote evaluation was
requested.
"""
def __init__(self, aggregator):
Stopper.__init__(self)
AggregatorAnnotator.__init__(self, aggregator)
self.aggregator=aggregator
def __call__(self, x, f, opt):
opt.stop=opt.stop or self.aggregator.allBelowOrAtZero()
# Reporter for reporting the results of aggregate function evaluation
# Reports details of every best-yet cf improvement
# One report per iterSpacing iterations without best-yet cf improvement
class AggregatorReporter(Reporter, AggregatorAnnotator):
"""
A subclass of the :class:`~pyopus.optimizer.base.Reporter` iterative
algorithm plugin class that reports the details of the last evaluated
aggregate function value obtained by the *aggregator* object. Uses
the :meth:`Aggregator.reportParameters` and
:meth:`Aggregator.formatResults` methods of *aggregator* for
obtaining the output that is printed at first iteration and every time
the aggregate function value decreases.
If *reportParameters* is ``True`` the report includes the corresponding
input parameter values.
This class is also an annotator that collects the results at remote
evaluation and copies them to the host where the remote evaluation was
requested.
"""
def __init__(self, aggregator, reportParameters=True):
Reporter.__init__(self)
AggregatorAnnotator.__init__(self, aggregator)
self.aggregator=aggregator
self.reportParameters=reportParameters
def __call__(self, x, f, opt):
# Print basic information for every iteration
Reporter.__call__(self, x, f, opt)
# Print details for first iteration and for every improvement
details=(opt.f is None) or (opt.niter==opt.bestIter)
# Print details. This requires an annotation either to be received or created.
if details and not self.quiet:
# Report details
# Print parameters
msg=self.aggregator.formatParameters(x)
print(msg)
# Print performance
msg=self.aggregator.formatResults()
print(msg)