def test_1D(self):
'''test a simple 1D linear interpolation'''
smin = np.array([0.])
smax = np.array([2.])
orders = np.array([3]) # ie. grid = [0,1,2]
grid = np.linspace(smin[0], smax[0], orders[0])
def f(x):
'function to interpolate'
return x**2
values = f(grid)
values = np.ascontiguousarray(np.atleast_2d(values))
eval_pts = np.linspace(smin[0], smax[0], 5)
eval_pts = np.ascontiguousarray(np.atleast_2d(eval_pts))
print(eval_pts)
interp_values = multilinear_interpolation(smin, smax, orders, values,
eval_pts)
print('interpolated values:')
print(interp_values)
interp_expected = np.array([0, 0.5, 1, 2.5,
4]) # linear interp. "by hand"
print('expected interpolated values:')
print(interp_expected)
assert_true(np.all(np.abs(interp_values - interp_expected) < 1e-10))
def test_1D(self):
"""test a simple 1D linear interpolation"""
smin = np.array([0.0])
smax = np.array([2.0])
orders = np.array([3]) # ie. grid = [0,1,2]
grid = np.linspace(smin[0], smax[0], orders[0])
def f(x):
"function to interpolate"
return x ** 2
values = f(grid)
values = np.ascontiguousarray(np.atleast_2d(values))
eval_pts = np.linspace(smin[0], smax[0], 5)
eval_pts = np.ascontiguousarray(np.atleast_2d(eval_pts))
print(eval_pts)
interp_values = multilinear_interpolation(smin, smax, orders, values, eval_pts)
print("interpolated values:")
print(interp_values)
interp_expected = np.array([0, 0.5, 1, 2.5, 4]) # linear interp. "by hand"
print("expected interpolated values:")
print(interp_expected)
assert_true(np.all(np.abs(interp_values - interp_expected) < 1e-10))
def __call__(self, *x_interp):
'''evaluate the interpolated function at coordinates `x_interp`
output shape is the shape of broadcasted coordinate inputs.
'''
assert len(x_interp) == self.ndim
# Prepare the interpolated coordinates array
x_mesh = np.broadcast_arrays(*x_interp)
shape = x_mesh[0].shape
x_stack = np.row_stack([x.astype(float).ravel() for x in x_mesh])
#
a = multilinear_interpolation(self._xmin, self._xmax, self._xshape,
self.values, x_stack)
a = a.reshape(shape)
return a
def __call__(self, *x_interp):
'''evaluate the interpolated function at coordinates `x_interp`
output shape is the shape of broadcasted coordinate inputs.
'''
assert len(x_interp) == self.ndim
# Prepare the interpolated coordinates array
x_mesh = np.broadcast_arrays(*x_interp)
shape = x_mesh[0].shape
x_stack = np.row_stack([x.astype(float).ravel() for x in x_mesh])
#
a = multilinear_interpolation(self._xmin, self._xmax, self._xshape,
self.values, x_stack)
a = a.reshape(shape)
return a
