'''
This file tests to make sure conversion from numpy -> uqtkarray does
not change the row-major (C contiguous) format of the original numpy array
Also, when converting form uqtkarray-> numpy we want to make sure that the
resulting numpy array is *only* row major (C contiguous)
'''
# create numpy matrix and show flags
a_np = np.array([[0, 2.00], [0.1, 1], [1, 5.0]])
print("flags for a_np to show whether C or F contiguous")
print(a_np.flags)
# get a uqtk array from a numpy array (memory is copied, not shared)
a_uqtk = numpy2uqtk(a_np)
print(
"\nflags for original numpy array to make sure it hasn't changed to F continguous after converting"
)
# verify that the original numpy array is only C contiguous
assert a_np.flags['F_CONTIGUOUS'] == False
assert a_np.flags['C_CONTIGUOUS'] == True
print("\nConvert uqtk array back to numpy array and make sure C contiguous")
b_np = uqtk2numpy(a_uqtk)
# test to make sure new numpy array is *only* C contiguous (row - major)
assert b_np.flags['F_CONTIGUOUS'] == False
assert b_np.flags['C_CONTIGUOUS'] == True
# test for the dot product
print(
self.__errbars_uqtk = uqtkarray.dblArray1D() # error bars for each weight
self.__nextbasis_uqtk = uqtkarray.dblArray1D() # if adaptive
self.__alpha_uqtk = uqtkarray.dblArray1D() # prior hyperparameter (1/gamma)
self.__lambda_p = uqtktools.new_doublep()
uqtktools.doublep_assign(self.__lambda_p,l_init)
self.__compiled = True
def leastsq(self,X,y):
'''
perform simple least squares based on the original
pc order.
'''
# convert input to uqtk arrays
self.__X_uqtk = numpy2uqtk(X)
self.__y_uqtk = numpy2uqtk(y)
# get vandermonde matrix w.r.t. original pc basis
self.__V_uqtk = uqtkarray.dblArray2D()
self.__pcmodel0 = uqtkpce.PCSet("NISPnoq",self.__mindex0_uqtk,self.pctype,0.0,1.0) # initiate
self.__pcmodel0.EvalBasisAtCustPts(self.__X_uqtk,self.__V_uqtk)
self.Vandermonde = uqtk2numpy(self.__V_uqtk)
self.__sol = np.linalg.lstsq(self.Vandermonde,y)
return self.__sol[0], self.__sol[1]
def fit0(self,X,y,tol=1e-8,sigsq=None,upit=0):
'''
Train bcs model coefficients with X and y data
X : 2d numpy of inputs/ feature data
y : 1d numpy array of labels/ outputs
# set 2d array to numpy array # make sure to pass asfortranarray y_np = random.randn(m,n) y.setnpdblArray(asfortranarray(y_np)) for i in range(m): for j in range(n): assert y[i,j] == y_np[i,j] ''' alternative using uqtk2numpy and numpy2uqtk ''' # test conversion from 1d numpy array to 1d uqtk array nn = 10 x1 = random.rand(nn) y1 = numpy2uqtk(x1) z1 = uqtk2numpy(y1) for i in range(nn): assert x1[i] == y1[i] # test conversion from 1d uqtk array to numpy for i in range(nn): assert z1[i] == x1[i] # test for conversion from 2d numpy to 2d uqtk nn = 10 mm = 5 X1 = random.rand(mm,nn) Y1 = numpy2uqtk(X1) Z1 = uqtk2numpy(Y1) for i in range(mm):
