def test_eval(self):
#
# 1D
#
# Define Kernel function
def k_fn(x,y,c = 2):
return x*y + c
k = Explicit(f=lambda x,y,c: x*y+c, parameters={'c':1},
n_variables=2, dim=1)
# Construct kernel, specifying parameter c
kernel = Kernel(k)
# Evaluation points
x = np.ones((11,1))
y = np.linspace(0,1,11)[:,None]
# Check accuracy
self.assertTrue(np.allclose(kernel.eval((x,y)), x*y+1))
# Define kernel with default parameters
k.set_parameters({'c':2})
kernel = Kernel(k)
# Check accuracy
self.assertTrue(np.allclose(kernel.eval((x,y)), x*y+2))
def test_set_parameters(self):
# Define explicit function
rule = lambda x, a: a * x[:, 0] - x[:, 1]
parms = [{'a': 1}, {'a': 2}]
f = Explicit(rule, parameters=parms, dim=2)
x = (1, 2)
self.assertTrue(np.allclose(f.eval(x), np.array([-1, 0])))
# Modify parameter
f.set_parameters({'a': 2}, pos=0)
self.assertTrue(np.allclose(f.eval(x), np.array([0, 0])))