def _compareBatchSelfAdjointEigRank2(self, x, use_gpu=False):
with self.test_session() as sess:
tf_eig = tf.batch_self_adjoint_eig(tf.constant(x))
tf_out = sess.run([tf_eig])[0]
dlist = x.shape
d = dlist[-2]
self.assertEqual(len(tf_eig.get_shape()), 2)
self.assertEqual([d+1, d], tf_eig.get_shape().dims[-2:])
self._testEigs(x, d, tf_out)
def _compareBatchSelfAdjointEigRank2(self, x, use_gpu=False):
with self.test_session() as sess:
tf_eig = tf.batch_self_adjoint_eig(tf.constant(x))
tf_out = sess.run([tf_eig])[0]
dlist = x.shape
d = dlist[-2]
self.assertEqual(len(tf_eig.get_shape()), 2)
self.assertEqual([d+1, d], tf_eig.get_shape().dims[-2:])
self._testEigs(x, d, tf_out)
def testWrongDimensions(self):
# The input to self_adjoint_eig should be 2-dimensional tensor.
scalar = tf.constant(1.)
with self.assertRaises(ValueError):
tf.self_adjoint_eig(scalar)
vector = tf.constant([1., 2.])
with self.assertRaises(ValueError):
tf.self_adjoint_eig(vector)
tensor = tf.constant([[[1., 2.], [3., 4.]], [[1., 2.], [3., 4.]]])
with self.assertRaises(ValueError):
tf.self_adjoint_eig(tensor)
# The input to batch_batch_self_adjoint_eig should be a tensor of
# at least rank 2.
scalar = tf.constant(1.)
with self.assertRaises(ValueError):
tf.batch_self_adjoint_eig(scalar)
vector = tf.constant([1., 2.])
with self.assertRaises(ValueError):
tf.batch_self_adjoint_eig(vector)
def testWrongDimensions(self):
# The input to self_adjoint_eig should be 2-dimensional tensor.
scalar = tf.constant(1.)
with self.assertRaises(ValueError):
tf.self_adjoint_eig(scalar)
vector = tf.constant([1., 2.])
with self.assertRaises(ValueError):
tf.self_adjoint_eig(vector)
tensor = tf.constant([[[1., 2.], [3., 4.]], [[1., 2.], [3., 4.]]])
with self.assertRaises(ValueError):
tf.self_adjoint_eig(tensor)
# The input to batch_batch_self_adjoint_eig should be a tensor of
# at least rank 2.
scalar = tf.constant(1.)
with self.assertRaises(ValueError):
tf.batch_self_adjoint_eig(scalar)
vector = tf.constant([1., 2.])
with self.assertRaises(ValueError):
tf.batch_self_adjoint_eig(vector)
def _compareBatchSelfAdjointEigRank3(self, x, use_gpu=False):
with self.test_session() as sess:
tf_eig = tf.batch_self_adjoint_eig(tf.constant(x))
tf_out = sess.run([tf_eig])[0]
dlist = x.shape
d = dlist[-2]
self.assertEqual([d+1, d], tf_eig.get_shape().dims[-2:])
# not testing the values.
self.assertEqual(dlist[0], tf_eig.get_shape().dims[0])
for i in xrange(dlist[0]):
self._testEigs(x[i], d, tf_out[i])
def _compareBatchSelfAdjointEigRank3(self, x, use_gpu=False):
with self.test_session() as sess:
tf_eig = tf.batch_self_adjoint_eig(tf.constant(x))
tf_out = sess.run([tf_eig])[0]
dlist = x.shape
d = dlist[-2]
self.assertEqual([d+1, d], tf_eig.get_shape().dims[-2:])
# not testing the values.
self.assertEqual(dlist[0], tf_eig.get_shape().dims[0])
for i in xrange(dlist[0]):
self._testEigs(x[i], d, tf_out[i])
# tf.cholesky
x = np.random.rand(10, 10)
z_cholesky = tf.cholesky(x)
# tf.batch_cholesky
batch_x = np.random.rand(10, 5, 5)
z_batch_cholesky = tf.batch_cholesky(x)
# tf.self_adjoint_eig
x = np.random.rand(10, 8)
z_self_adjoint_eig = tf.self_adjoint_eig(x)
# tf.batch_self_adjoint_eig
batch_x = np.random.rand(10, 8, 5)
z_batch_self_adjoint_eig = tf.batch_self_adjoint_eig(batch_x)
with tf.Session() as sess:
print "tf.diag"
print sess.run(z_diag)
print "tf.transpose"
print sess.run(z_transpose)
print "tf.matmul"
print sess.run(z_matmul)
print "tf.batch_matmul"
print sess.run(z_batch_matmul)
print "tf.matrix_determinant"
def test_BatchSelfAdjointEig(self):
t = tf.batch_self_adjoint_eig(np.array(3 * [3, 2, 2, 1]).reshape(3, 2, 2).astype("float32"))
self.check(t, ndigits=4, stats=True)
