def __trainLocal__(self,featureVals,targetVals):
"""
Because HDMR rom is a collection of sub-roms, we call sub-rom "train" to do what we need it do.
@ In, featureVals, np.array, training feature values
@ In, targetVals, np.array, training target values
@ Out, None
"""
if not self.initialized:
self.raiseAnError(RuntimeError,'ROM has not yet been initialized! Has the Sampler associated with this ROM been used?')
ft={}
self.refSoln = {key:dict({}) for key in self.target}
for i in range(len(featureVals)):
ft[tuple(featureVals[i])]=targetVals[i,:]
#get the reference case
self.refpt = tuple(self.__fillPointWithRef((),[]))
for cnt, target in enumerate(self.target):
self.refSoln[target] = ft[self.refpt][cnt]
for combo,rom in self.ROMs.items():
subtdict = {key:list([]) for key in self.target}
for c in combo:
subtdict[c]=[]
SG = rom.sparseGrid
fvals=np.zeros([len(SG),len(combo)])
tvals=np.zeros((len(SG),len(self.target)))
for i in range(len(SG)):
getpt=tuple(self.__fillPointWithRef(combo,SG[i][0]))
#the 1e-10 is to be consistent with RAVEN's CSV print precision
tvals[i,:] = ft[tuple(mathUtils.NDInArray(np.array(list(ft.keys())),getpt,tol=1e-10)[2])]
for fp,fpt in enumerate(SG[i][0]):
fvals[i][fp] = fpt
for i,c in enumerate(combo):
subtdict[c] = fvals[:,i]
for cnt, target in enumerate(self.target):
subtdict[target] = tvals[:,cnt]
rom.train(subtdict)
#make ordered list of combos for use later
maxLevel = max(list(len(combo) for combo in self.ROMs.keys()))
self.combos = []
for i in range(maxLevel+1):
self.combos.append([])
for combo in self.ROMs.keys():
self.combos[len(combo)].append(combo)
#list of term objects
self.terms = {():[]} # each entry will look like 'x1,x2':('x1','x2'), missing the reference entry
for l in range(1,maxLevel+1):
for romName in self.combos[l]:
self.terms[romName] = []
# add subroms -> does this get referenece case, too?
for key in self.terms.keys():
if set(key).issubset(set(romName)) and key!=romName:
self.terms[romName].append(key)
#reduce terms
self.reducedTerms = {}
for term in self.terms.keys():
self._collectTerms(term,self.reducedTerms)
#remove zero entries
self._removeZeroTerms(self.reducedTerms)
self.amITrained = True
checkTrue('compareFloats moderate OoM match',mathUtils.compareFloats(3.141592,3.141593,tol=1e-6),True)
checkTrue('compareFloats moderate OoM mismatch',mathUtils.compareFloats(3.141592,3.141593,tol=1e-8),False)
#small order of magnitude
checkTrue('compareFloats small OoM match',mathUtils.compareFloats(3.141592e-15,3.141593e-15,tol=1e-6),True)
checkTrue('compareFloats small OoM mismatch',mathUtils.compareFloats(3.141592e-15,3.141593e-15,tol=1e-8),False)
#small order of magnitude
checkTrue('compareFloats large OoM match',mathUtils.compareFloats(3.141592e15,3.141593e15,tol=1e-6),True)
checkTrue('compareFloats large OoM mismatch',mathUtils.compareFloats(3.141592e15,3.141593e15,tol=1e-8),False)
### check "NDinArray"
points = np.array([(0.61259532,0.27325707,0.81182424),
(0.54608679,0.82470626,0.39170769)])
findSmall = (0.55,0.82,0.39)
findLarge = (0.61259532123,0.27325707123,0.81182423999)
found,idx,entry = mathUtils.NDInArray(points,findSmall,tol=1e-2)
checkAnswer('NDInArray %s found' %str(findSmall),int(found),1)
checkAnswer('NDInArray %s idx' %str(findSmall),idx,1)
checkArray('NDInArray %s entry' %str(findSmall),entry,points[1])
found,idx,entry = mathUtils.NDInArray(points,findSmall,tol=1e-3)
checkAnswer('NDInArray %s not found' %str(findSmall),int(found),0)
checkType('NDInArray %s no idx' %str(findSmall),idx,None)
checkType('NDInArray %s no entry' %str(findSmall),entry,None)
found,idx,entry = mathUtils.NDInArray(points,findLarge,tol=1e-8)
checkAnswer('NDInArray %s found' %str(findLarge),int(found),1)
checkAnswer('NDInArray %s idx' %str(findLarge),idx,0)
checkArray('NDInArray %s entry' %str(findLarge),entry,points[0])
### check "normalizationFactors"
zeroList = [0,0,0,0,0]
fourList = [4,4,4,4,4]