def __init__(self, dim, numTimeSteps=451, gridResolution=16):
self.dim = dim
self.out = numTimeSteps
self.gridResolution = gridResolution
self.okushiriStorage = okushiriStorage(dim, numTimeSteps,
self.gridResolution)
self.pdfs = pysgpp.DistributionsVector()
for d in range(self.dim):
self.pdfs.push_back(pysgpp.DistributionUniform(0.0, 2.0))
def __init__(self,
dim,
numTimeSteps=451,
gridResolution=16,
normalization=1,
residual=0,
wave_type='bumps',
distribution='normal',
minimum_allowed_height=1e-5):
self.dim = dim
self.out = numTimeSteps
self.gridResolution = gridResolution # 16/128
self.normalization = normalization
self.residual = residual
self.wave_type = wave_type
self.okushiriStorage = okushiriStorage(dim, numTimeSteps,
self.gridResolution,
self.normalization,
self.residual, self.wave_type)
self.pdfs = pysgpp.DistributionsVector()
self.distribution = distribution
self.minimum_allowed_height = minimum_allowed_height
if wave_type == 'bumps':
for d in range(self.dim):
if self.distribution == 'uniform':
self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
elif self.distribution == 'normal':
self.pdfs.push_back(
pysgpp.DistributionTruncNormal(1.0, 0.125, 0.5, 1.5))
# self.pdfs.push_back(pysgpp.DistributionNormal(1.0, 0.125))
# elif wave_type == 'shape':
# # TODO REPLACE THESE WITH NORMAL DISTRIBUTIONS WHEN CALCULATING STOCHASTIC MOMENTS
# self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
# self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
# self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
# self.pdfs.push_back(pysgpp.DistributionUniform(1.0, 2.0))
# self.pdfs.push_back(pysgpp.DistributionUniform(1.0, 2.0))
# self.pdfs.push_back(pysgpp.DistributionUniform(1.0, 2.0))
elif wave_type == 'original':
for _ in range(self.dim):
self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
def __init__(self, dim, normalization=1, residual=1):
self.dim = dim
self.normalization = normalization
self.residual = residual
self.out = 451
self.pdfs = pysgpp.DistributionsVector()
for d in range(dim):
self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
data = np.loadtxt(
'/home/rehmemk/git/anugasgpp/Okushiri/data/boundary_wave_original.txt',
skiprows=1)
self.t = data[:, 0]
self.y = data[:, 1]
self.energy = np.trapz(self.y**2, self.t)
def __init__(self,
dim,
reallb,
realub,
numTimeSteps=451,
gridResolution=16,
normalization=1,
residual=0):
self.dim = dim
self.reallb = reallb
self.realub = realub
self.gridResolution = gridResolution
self.numTimeSteps = numTimeSteps
self.normalization = normalization
self.residual = residual
self.okushiriStorage = okushiriStorage(dim, self.numTimeSteps,
self.gridResolution,
self.normalization,
self.residual)
self.pdfs = pysgpp.DistributionsVector()
for _ in range(dim):
self.pdfs.push_back(pysgpp.DistributionUniform(0.0, 1.0))
def __init__(self, gridResolution=16, normalization=1):
self.dim = 6
self.gridResolution = gridResolution
self.normalization = normalization
self.numTimeSteps = 451
residual = 0
self.wave_type = 'shape'
self.okushiriStorage = okushiriStorage(self.dim, self.numTimeSteps,
self.gridResolution,
self.normalization, residual,
self.wave_type)
self.pdfs = pysgpp.DistributionsVector()
# TODO REPLACE THESE WITH NORMAL DISTRIBUTIONS WHEN CALCULATING STOCHASTIC MOMENTS
self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
self.pdfs.push_back(pysgpp.DistributionUniform(0.5, 1.5))
self.pdfs.push_back(pysgpp.DistributionUniform(1.0, 2.0))
self.pdfs.push_back(pysgpp.DistributionUniform(1.0, 2.0))
self.pdfs.push_back(pysgpp.DistributionUniform(1.0, 2.0))
reSurf.surplusAdaptive(numPoints, initialLevel, numRefine, verbose)
# Now the surrogate has been created and we can use it to calculate various
# quantities of interest
# evaluate the surrogate
evalPoint = pysgpp.DataVector([0.3, 0.6])
print(f'reSurf: {reSurf.eval(evalPoint)} truth: {objFunc.eval(evalPoint)}')
# evaluate the surrogate's gradient
gradient = pysgpp.DataVector(dim, 0)
reSurf.evalGradient(evalPoint, gradient)
print(f'gradient: {gradient.toString()}')
# integrate the surrogate
print(f'integral: {reSurf.getIntegral()}')
# calculate stochastic moments
# first specify distibutions
pdfs = pysgpp.DistributionsVector()
pdfs.push_back(pysgpp.DistributionNormal(0.5, 0.1))
pdfs.push_back(pysgpp.DistributionUniform(0, 1))
# set quadrature order appropriate for the distributions and desired accuracy
quadOrder = 15
print(f'mean: {reSurf.getMean(pdfs,quadOrder)}')
v = reSurf.getVariance(pdfs, quadOrder)
print(f'variance: {v[0]}')
# optimize the surrogate
print(f'Optimum: {reSurf.optimize()}')