def __init__(self,messageHandler,**kwargs):
"""
PolyExponential constructor
@ In, messageHandler, MessageHandler.MessageUser, a MessageHandler object in charge of raising errors,
and printing messages
@ In, kwargs, dict, an arbitrary dictionary of keywords and values
"""
supervisedLearning.__init__(self,messageHandler,**kwargs)
self.availCoeffRegressors = ['nearest','poly','spline'] # avail coeff regressors
self.printTag = 'PolyExponential' # Print tag
self.pivotParameterID = kwargs.get("pivotParameter","time") # Pivot parameter ID
self._dynamicHandling = True # This ROM is able to manage the time-series on its own
self.polyExpParams = {} # poly exponential options' container
self.polyExpParams['expTerms'] = int(kwargs.get('numberExpTerms',3)) # the number of exponential terms
self.polyExpParams['coeffRegressor'] = kwargs.get('coeffRegressor','spline').lower() # which regressor to use for interpolating the coefficient
self.polyExpParams['polyOrder'] = int(kwargs.get('polyOrder',3)) # the polynomial order
self.polyExpParams['tol'] = float(kwargs.get('tol',0.001)) # optimization tolerance
self.polyExpParams['maxNumberIter'] = int(kwargs.get('maxNumberIter',5000)) # maximum number of iterations in optimization
self.aij = None # a_ij coefficients of the exponential terms {'target1':ndarray(nsamples, self.polyExpParams['expTerms']),'target2',ndarray,etc}
self.bij = None # b_ij coefficients of the exponent of the exponential terms {'target1':ndarray(nsamples, self.polyExpParams['expTerms']),'target2',ndarray,etc}
self.model = None # the surrogate model itself {'target1':model,'target2':model, etc.}
# check if the pivotParameter is among the targetValues
if self.pivotParameterID not in self.target:
self.raiseAnError(IOError,"The pivotParameter "+self.pivotParameterID+" must be part of the Target space!")
def __init__(self, **kwargs):
"""
DMD constructor
@ In, kwargs, dict, an arbitrary dictionary of keywords and values
"""
supervisedLearning.__init__(self, **kwargs)
self.availDmdAlgorithms = [
'dmd', 'hodmd'
] # available dmd types: basic dmd and high order dmd
self.dmdParams = {} # dmd settings container
self.printTag = 'DMD' # print tag
self.pivotParameterID = kwargs.get("pivotParameter",
"time") # pivot parameter
self._dynamicHandling = True # This ROM is able to manage the time-series on its own. No need for special treatment outside
self.dmdParams['rankSVD'] = kwargs.get(
'rankSVD', None
) # -1 no truncation, 0 optimal rank is computed, >1 truncation rank
self.dmdParams['energyRankSVD'] = kwargs.get(
'energyRankSVD', None
) # 0.0 < float < 1.0, computed rank is the number of the biggest sv needed to reach the energy identified by "energyRankSVD"
self.dmdParams['rankTLSQ'] = kwargs.get(
'rankTLSQ', None) # truncation rank for total least square
self.dmdParams['exactModes'] = kwargs.get(
'exactModes', True
) # True if the exact modes need to be computed (eigs and eigvs), otherwise the projected ones (using the left-singular matrix)
self.dmdParams['optimized'] = kwargs.get(
'optimized', False
) # amplitudes computed minimizing the error between the mods and all the timesteps (True) or 1st timestep only (False)
self.dmdParams['dmdType'] = kwargs.get(
'dmdType', 'dmd'
) # the dmd type to be applied. Currently we support dmd and hdmd (high order dmd)
# variables filled up in the training stages
self._amplitudes = {
} # {'target1': vector of amplitudes,'target2':vector of amplitudes, etc.}
self._eigs = {
} # {'target1': vector of eigenvalues,'target2':vector of eigenvalues, etc.}
self._modes = {
} # {'target1': matrix of dynamic modes,'target2':matrix of dynamic modes, etc.}
self.__Atilde = {
} # {'target1': matrix of lowrank operator from the SVD,'target2':matrix of lowrank operator from the SVD, etc.}
self.pivotValues = None # pivot values (e.g. time)
self.KDTreeFinder = None # kdtree weighting model
self.timeScales = {
} # time-scales (training and dmd). {'training' and 'dmd':{t0:float,'dt':float,'intervals':int}}
# some checks
if self.dmdParams['rankSVD'] is not None and self.dmdParams[
'energyRankSVD'] is not None:
self.raiseAWarning(
'Both "rankSVD" and "energyRankSVD" have been inputted. "energyRankSVD" is predominant and will be used!'
)
if self.dmdParams['dmdType'] not in self.availDmdAlgorithms:
self.raiseAnError(
IOError, 'dmdType(s) available are "' +
', '.join(self.availDmdAlgorithms) + '"!')
# check if the pivotParameter is among the targetValues
if self.pivotParameterID not in self.target:
self.raiseAnError(
IOError, "The pivotParameter " + self.pivotParameterID +
" must be part of the Target space!")
