def test_RecurrentConnection_forward_pass_single_samples(): rc = RecurrentConnection(4) theta = 2 * np.eye(4).flatten() for x in X: x = np.atleast_2d(x) # for single samples recurrence should not do anything assert_equal(rc.forward_pass(theta, x), x)
def test_RecurrentConnection_forward_pass_multi_sample(): fc = RecurrentConnection(4) theta = np.eye(4).flatten() X_summed = np.array([[0, 0, 0, 1], [1, 0, 0, 2],[1, 1, 0, 3],[1, 1, 1, 4],[2, 2, 1, 5]]) assert_equal(fc.forward_pass(theta, X), X_summed)
def test_RecurrentConnection_dimensions(): rc = RecurrentConnection(5) assert_equal(rc.input_dim, 5) assert_equal(rc.output_dim, 5) assert_equal(rc.get_param_dim(), 5**2)
def test_RecurrentConnection_forward_pass_multi_sample(): fc = RecurrentConnection(4) theta = np.eye(4).flatten() X_summed = np.array([[0, 0, 0, 1], [1, 0, 0, 2], [1, 1, 0, 3], [1, 1, 1, 4], [2, 2, 1, 5]]) assert_equal(fc.forward_pass(theta, X), X_summed)