def test_simplest_factorize(self):
print()
model = benchmark_model.build_tf_model()
ops = utility.get_train_ops()
layers = utility.zip_layer(inputs=model.inputs, ops=ops)
hidden = layers[-1].output
last_weights = layers[-1].kernel
tf_u, tf_v = semi_nmf(model.labels,
hidden,
last_weights,
use_tf=True,
use_bias=False,
num_iters=3)
_old_local_loss = losses.frobenius_norm(model.labels,
hidden @ last_weights)
_new_local_loss = losses.frobenius_norm(model.labels, tf_u @ tf_v)
x, y = benchmark_model.build_data(batch_size, label_size)
init = tf.global_variables_initializer()
with self.test_session() as sess:
sess.run(init)
old_local_loss, new_local_loss = sess.run(
[_old_local_loss, _new_local_loss],
feed_dict={
model.inputs: x,
model.labels: y,
})
self.assertGreater(old_local_loss, new_local_loss)
print("old {} new {}".format(old_local_loss, new_local_loss))
def build_tf_hitachi_simple_model(batch_size,
feature_size=3,
label_size=4,
use_bias=False,
activation=None):
inputs = tf.placeholder(tf.float64, (batch_size, feature_size),
name='inputs')
labels = tf.placeholder(tf.float64, (batch_size, label_size),
name='labels')
x = tf.layers.dense(inputs, 1000, activation=activation, use_bias=use_bias)
x = tf.layers.dense(x, 500, activation=activation, use_bias=use_bias)
outputs = tf.layers.dense(x,
label_size,
activation=None,
use_bias=use_bias)
losses = frobenius_norm(labels, outputs)
loss = tf.reduce_mean(losses)
mse = loss / tf.sqrt(tf.constant(batch_size,
tf.float64)) # teached from SAKURAI
# mse_losses = losses / tf.sqrt(tf.constant(batch_size, tf.float64)) # teached from SAKURAI
mse_losses = losses
return agents.tools.AttrDict(
inputs=inputs,
outputs=outputs,
labels=labels,
loss=loss,
mse=mse,
mse_losses=mse_losses,
)
def build_tf_hitachi_model(batch_size,
feature_size=3,
label_size=4,
use_bias=False,
activation=None):
num_layers = 7
units = 128
inputs = tf.placeholder(tf.float64, (batch_size, feature_size),
name='inputs')
labels = tf.placeholder(tf.float64, (batch_size, label_size),
name='labels')
x = inputs
for i in range(num_layers):
x = tf.layers.dense(x,
units=units,
activation=activation,
use_bias=use_bias)
outputs = tf.layers.dense(x,
label_size,
activation=None,
use_bias=use_bias)
losses = frobenius_norm(labels, outputs)
loss = tf.reduce_mean(losses)
return agents.tools.AttrDict(
inputs=inputs,
outputs=outputs,
labels=labels,
loss=loss,
)
def build_tf_one_hot_model(batch_size,
shape=784,
use_bias=False,
activation=None,
use_softmax=False):
inputs = tf.placeholder(tf.float64, (batch_size, shape), name='inputs')
labels = tf.placeholder(tf.float64, (batch_size, 10), name='labels')
activation = None or activation
x = tf.layers.dense(inputs, 1000, activation=activation, use_bias=use_bias)
x = tf.layers.dense(x, 500, activation=activation, use_bias=use_bias)
outputs = tf.layers.dense(x, 10, activation=None, use_bias=use_bias)
losses = frobenius_norm(labels, outputs)
frob_norm = tf.reduce_mean(losses)
other_losses = tf.nn.softmax_cross_entropy_with_logits_v2(labels=labels,
logits=outputs)
cross_entropy = tf.reduce_mean(other_losses)
correct_prediction = tf.equal(tf.argmax(labels, 1), tf.argmax(outputs, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) * 100.
return agents.tools.AttrDict(
inputs=inputs,
outputs=outputs,
labels=labels,
frob_norm=frob_norm,
cross_entropy=cross_entropy,
accuracy=accuracy,
)
def test_u_neg_nonlin_semi_nmf(self):
mat_file = python_path.joinpath('./u_neg.mat').as_posix()
auv = sio.loadmat(mat_file)
a, u, v = auv['a'], auv['u'], auv['v']
old_loss = np_frobenius_norm(a, u @ v)
a_ph = tf.placeholder(tf.float64, shape=a.shape)
u_ph = tf.placeholder(tf.float64, shape=u.shape)
v_ph = tf.placeholder(tf.float64, shape=v.shape)
tf_u, tf_v = tf.py_func(nonlin_semi_nmf, [a_ph, u_ph, v_ph],
[tf.float64, tf.float64])
tf_loss = frobenius_norm(a_ph, tf.nn.relu(tf_u @ tf_v))
init = tf.global_variables_initializer()
with tf.Session() as sess:
init.run()
start_time = time.time()
_u, _v, new_loss = sess.run([tf_u, tf_v, tf_loss],
feed_dict={
a_ph: a,
u_ph: u,
v_ph: v
})
end_time = time.time()
duration = end_time - start_time
assert a.shape == (_u @ _v).shape
assert new_loss < old_loss, "new loss should be less than old loss."
print('solve Nonlinear semi-NMF\n\t'
'old loss {0}\n\t'
'new loss {1}\n\t'
'process duration {2}'.format(old_loss, new_loss, duration))
def test_tf_not_calc_v_nonlin_semi_nmf(self):
auv = sio.loadmat(mat_file)
a, u, v = auv['a'], auv['u'], auv['v']
old_loss = np_frobenius_norm(a, u @ v)
# [1000, 500]
a_ph = tf.placeholder(tf.float64, shape=a.shape)
# [1000, 201]
u_ph = tf.placeholder(tf.float64, shape=u.shape)
# [200, 500]
v_ph = tf.placeholder(tf.float64, shape=v.shape)
tf_u, tf_v = nonlin_semi_nmf(a_ph, u_ph, v_ph, use_tf=True, use_bias=False, num_calc_v=0, num_calc_u=1)
tf_loss = frobenius_norm(a_ph, tf.nn.relu(tf.matmul(tf_u, tf_v)))
init = tf.global_variables_initializer()
with tf.Session() as sess:
init.run()
start_time = time.time()
_u, _v, new_loss = sess.run([tf_u, tf_v, tf_loss], feed_dict={a_ph: a, u_ph: u, v_ph: v})
end_time = time.time()
duration = end_time - start_time
assert a.shape == (_u @ _v).shape
assert new_loss < old_loss, "new loss should be less than old loss."
print_format('TensorFlow', 'Nonlinear semi-NMF(NOT CALCLATE v)', a, u, v, old_loss, new_loss, duration)
def test_tf_vanilla_semi_nmf(self):
a = np.random.uniform(-1., 1., size=(100, 100))
u = np.random.uniform(0., 1., size=(100, 300))
v = np.random.uniform(-1., 1., size=(300, 100))
old_loss = np_frobenius_norm(a, u @ v)
# [1000, 500]
a_ph = tf.placeholder(tf.float64, shape=a.shape)
# [1000, 201]
u_ph = tf.placeholder(tf.float64, shape=u.shape)
# [200, 500]
v_ph = tf.placeholder(tf.float64, shape=v.shape)
tf_u, tf_v = semi_nmf(a_ph, u_ph, v_ph, use_tf=True, use_bias=False)
tf_loss = frobenius_norm(a_ph, tf.matmul(tf_u, tf_v))
init = tf.global_variables_initializer()
with tf.Session() as sess:
init.run()
start_time = time.time()
_u, _v, new_loss = sess.run([tf_u, tf_v, tf_loss], feed_dict={a_ph: a, u_ph: u, v_ph: v})
assert np.min(_u) > 0, np.min(_u)
end_time = time.time()
duration = end_time - start_time
assert a.shape == (_u @ _v).shape
assert new_loss < old_loss, "new loss should be less than old loss."
print('\n[TensorFlow]Solve semi-NMF\n\t'
'old loss {0}\n\t'
'new loss {1}\n\t'
'process duration {2}'.format(old_loss, new_loss, duration))
print_format('TensorFlow', 'semi-NMF', a, u, v, old_loss, new_loss, duration)
def build_rnn_mnist(batch_size, use_bias=False, activation=None):
time_steps = 28
num_features = 28
inputs = tf.keras.layers.Input((time_steps, num_features), batch_size=batch_size, dtype=tf.float64, name='inputs')
labels = tf.placeholder(tf.float64, (batch_size, 10), name='labels')
activation = None or activation
x = tf.keras.layers.SimpleRNN(100, use_bias=use_bias, activation=activation, return_sequences=True)(inputs)
x = x[:, -1, :]
# x = tf.keras.layers.SimpleRNN(100, use_bias=use_bias, activation=activation)(inputs)
# print(x)
# print(search(x))
outputs = tf.layers.dense(x, 10, activation=None, use_bias=use_bias)
losses = frobenius_norm(labels, outputs)
frob_norm = tf.reduce_mean(losses)
other_losses = tf.nn.softmax_cross_entropy_with_logits_v2(labels=labels, logits=outputs)
cross_entropy = tf.reduce_mean(other_losses)
correct_prediction = tf.equal(tf.argmax(labels, 1), tf.argmax(outputs, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) * 100.
return agents.tools.AttrDict(inputs=inputs,
outputs=outputs,
labels=labels,
frob_norm=frob_norm,
cross_entropy=cross_entropy,
accuracy=accuracy,
)
def test_tf_vanilla_semi_nmf(self):
auv = sio.loadmat(mat_file)
a, u, v = auv['a'], auv['u'], auv['v']
old_loss = np_frobenius_norm(a, u @ v)
# [1000, 500]
a_ph = tf.placeholder(tf.float64, shape=a.shape)
# [1000, 201]
u_ph = tf.placeholder(tf.float64, shape=u.shape)
# [200, 500]
v_ph = tf.placeholder(tf.float64, shape=v.shape)
tf_u, tf_v = semi_nmf(a_ph, u_ph, v_ph, use_tf=True, use_bias=False)
tf_loss = frobenius_norm(a_ph, tf.matmul(tf_u, tf_v))
init = tf.global_variables_initializer()
with tf.Session() as sess:
init.run()
start_time = time.time()
_u, _v, new_loss = sess.run([tf_u, tf_v, tf_loss],
feed_dict={
a_ph: a,
u_ph: u,
v_ph: v
})
end_time = time.time()
duration = end_time - start_time
assert a.shape == (_u @ _v).shape
assert new_loss < old_loss, "new loss should be less than old loss."
print('\n[TensorFlow]Solve semi-NMF\n\t'
'old loss {0}\n\t'
'new loss {1}\n\t'
'process duration {2}'.format(old_loss, new_loss, duration))
def build_tf_model():
batch_size = 3000
label_size = 1
inputs = tf.placeholder(tf.float64, (batch_size, 784), name='inputs')
labels = tf.placeholder(tf.float64, (batch_size, label_size),
name='labels')
x = tf.layers.dense(inputs, 100, activation=tf.nn.relu, use_bias=True)
x = tf.layers.dense(x, 50, use_bias=False, activation=tf.nn.relu)
outputs = tf.layers.dense(x, label_size, activation=None, use_bias=True)
losses = frobenius_norm(labels, outputs)
loss = tf.reduce_mean(losses)
correct_prediction = tf.equal(tf.cast(labels, tf.int32),
tf.cast(outputs, tf.int32))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) * 100.
tf.summary.scalar('accuracy', accuracy)
return agents.tools.AttrDict(
inputs=inputs,
outputs=outputs,
labels=labels,
loss=loss,
accuracy=accuracy,
)