def random_instance(self, *args, **kwargs):
U = util.orthogonal(self.d)
D = np.random.rand(self.d)
M = U.dot(np.diag(D)).dot(U.T)
L = np.linalg.cholesky(M)
return self.R, self.matrix_to_polys(M), L
def random_instance(self, *args, **kwargs):
U = util.orthogonal(self.d)
D = np.random.rand(self.d)
M = U.dot(np.diag(D)).dot(U.T)
L = np.linalg.cholesky(M)
return self.R, self.matrix_to_polys(M), L
def random_instance(self, *args, **kwargs):
U = util.orthogonal(self.d)
D = np.random.rand(self.d)
M = U.dot(np.diag(D)).dot(U.T)
print D, M
return M, self.R, self.matrix_to_polys(M), D
def random_instance(self, *args, **kwargs):
U = util.orthogonal(self.d)
D = np.random.rand(self.d)
M = U.dot(np.diag(D)).dot(U.T)
print D, M
return M, self.R, self.matrix_to_polys(M), D
def test_candecomp():
"""
Test if it works
"""
d = 3
pi = rand(d)
A = orthogonal(3)
T = ten.ktensor( [A, A, A], pi ).totensor()
pi_, A_, B_, C_ = candecomp( T, d )
T_ = ten.ktensor( [A_, B_, C_], pi_ ).totensor()
pi_ = match_columns_sign(np.atleast_2d(pi_), np.atleast_2d(pi)).flatten()
A_ = match_columns_sign(A_, A)
assert np.allclose(pi, pi_)
assert np.allclose(A, A_)
assert np.allclose( T, T_ )
def test_candecomp():
"""
Test if it works
"""
d = 3
pi = rand(d)
A = orthogonal(3)
T = ten.ktensor([A, A, A], pi).totensor()
pi_, A_, B_, C_ = candecomp(T, d)
T_ = ten.ktensor([A_, B_, C_], pi_).totensor()
pi_ = match_columns_sign(np.atleast_2d(pi_), np.atleast_2d(pi)).flatten()
A_ = match_columns_sign(A_, A)
assert np.allclose(pi, pi_)
assert np.allclose(A, A_)
assert np.allclose(T, T_)
inputs = ordict(features=T.matrix("features"),
targets=T.ivector("targets"))
x, y = inputs["features"], inputs["targets"]
theano.config.compute_test_value = "warn"
x.tag.test_value = np.random.random((11, 784)).astype(theano.config.floatX)
y.tag.test_value = np.random.random_integers(low=0, high=9, size=(11,)).astype(np.int32)
# move time axis before batch axis
x = x.T
# (time, batch, features)
x = x.reshape((x.shape[0], x.shape[1], 1))
Wx = theano.shared(util.orthogonal((1, nhidden)), name="Wx")
bx = theano.shared(np.zeros((nhidden,), dtype=theano.config.floatX), name="bx")
Wy = theano.shared(util.orthogonal((nhidden, nclasses)), name="Wy")
by = theano.shared(np.zeros((nclasses,), dtype=theano.config.floatX), name="by")
Wh = theano.shared(0.9*np.eye(nhidden, dtype=theano.config.floatX), name="Wh")
h0 = theano.shared(np.zeros((nhidden,), dtype=theano.config.floatX), name="h0")
parameters = ordict((var.name, var) for var in [Wx, bx, Wy, by, Wh, h0])
taps = [-1] + [-m for m in args.skips]
if args.instance_dependent:
Wskip = theano.shared(np.zeros((nhidden, len(taps),), dtype=theano.config.floatX), name="Wskip")
bskip = theano.shared(np.zeros((len(taps),), dtype=theano.config.floatX), name="bskip")
parameters["Wskip"] = Wskip
parameters["bskip"] = bskip
