def test_convolve1d_3():
npr.seed(3)
for ii in xrange(NUM_TRIALS):
np_A = npr.randn(5, 50)
np_B = npr.randn(6 * 5, 4)
A = kayak.Parameter(np_A)
B = kayak.Parameter(np_B)
C = kayak.Convolve1d(A, B, ncolors=5)
assert_equals(C.value.shape, (5, (10 - 6 + 1) * 4))
def test_convolve1d_1():
npr.seed(3)
for ii in xrange(NUM_TRIALS):
np_A = npr.randn(5, 6)
np_B = npr.randn(6, 7)
A = kayak.Parameter(np_A)
B = kayak.Parameter(np_B)
C = kayak.Convolve1d(A, B, ncolors=1)
# If the filters are the same size as the data
assert C.value.shape == (5, 7)
def test_convolve1d_grad_1():
npr.seed(3)
for ii in xrange(NUM_TRIALS):
np_A = npr.randn(5, 6)
np_B = npr.randn(6, 7)
A = kayak.Parameter(np_A)
B = kayak.Parameter(np_B)
C = kayak.Convolve1d(A, B)
D = kayak.MatSum(C)
D.value
assert_equals(D.grad(A).shape, (5, 6))
assert_equals(D.grad(B).shape, (6, 7))
assert_less(kayak.util.checkgrad(A, D), MAX_GRAD_DIFF)
assert_less(kayak.util.checkgrad(B, D), MAX_GRAD_DIFF)