def test_ReflectedKernel():
kernel = lambda pt : np.abs(pt)
testkernel = testmod.ReflectedKernel(kernel)
assert( testkernel(5) == 10 )
np.testing.assert_allclose(testkernel([1,2,3]), [2,4,6])
# 2 (or 3 etc.) dim kernel only
testkernel = testmod.ReflectedKernel(lambda pt : np.abs(pt[0]))
np.testing.assert_allclose(testkernel([[1,2,3],[4,5,6]]), [2,4,6])
testkernel = testmod.ReflectedKernel(lambda pt : pt[0] * (pt[0]>=0))
np.testing.assert_allclose(testkernel([[1,2,3],[4,5,6]]), [1,2,3])
testkernel = testmod.ReflectedKernel(lambda pt : pt[0] * (pt[0]>=0), reflected_axis=1)
np.testing.assert_allclose(testkernel([[1,2,3],[4,5,6]]), [2,4,6])
def compute_space_marginal(ker):
"""Return the kernel resulting from integrating out all but the first
coordinate. Handles the case when a cutoff is applied, but does not apply
the cutoff in the calculation!
:param ker: Any of the kernels returned by the providers in this module.
"""
already_reflected_axis = None
if isinstance(ker, kernels.ReflectedKernel):
already_reflected_axis = ker.reflected_axis
ker = ker.delegate
if isinstance(ker, _WrapWithCutoff):
ker = ker._ker
k = kernels.marginalise_gaussian_kernel(ker, 0)
if already_reflected_axis is not None and already_reflected_axis > 0:
return kernels.ReflectedKernel(k, already_reflected_axis - 1)
return k
def trigger_opt(self):
_, trigs = self._sample_points()
ker = self.trigger_provider(trigs, None)
return kernels.ReflectedKernel(ker, 0)
def trigger_opt(self):
ker = self.trigger_provider(*self.data_for_trigger_opt())
return kernels.ReflectedKernel(ker, 0)
def ker_opt(self):
ker = self.kernel_provider(*self.data_for_ker_opt())
return kernels.ReflectedKernel(ker, 0)