def classifyMammograms():
X, y = readData()
X = numpy.array(X)
y = numpy.array(y)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.20,
random_state=123,
shuffle=True)
y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
model = Sequential()
model.add(
Dense(128,
activation='relu',
input_dim=len(X[0]),
kernel_regularizer='l2',
kernel_initializer=he_normal(seed=None)))
model.add(Dense(128))
model.add(Dense(64, activation='relu'))
model.add(Dense(64))
model.add(Dense(32, activation='relu'))
model.add(Dense(32))
model.add(Dense(16, activation='relu'))
model.add(Dense(16))
model.add(Dense(8, activation='relu'))
model.add(Dense(4))
model.add(Dense(2, activation='sigmoid'))
model.compile(optimizer='adam',
loss='categorical_crossentropy',
metrics=['accuracy'])
filepath = "model.hdf5"
checkpoint = ModelCheckpoint(filepath,
monitor='loss',
verbose=1,
save_best_only=True,
mode='min')
earlystopping = EarlyStopping(monitor='loss',
patience=20,
restore_best_weights=True)
desired_callbacks = [checkpoint, earlystopping]
# history = model.fit(X_train, y_train, validation_split=0.2, batch_size=50, epochs=1000, shuffle=True,
# callbacks=desired_callbacks)
# plt.plot(history.history['loss'])
# plt.plot(history.history['val_loss'])
# plt.title('model loss')
# plt.ylabel('loss')
# plt.xlabel('epoch')
# plt.legend(['train', 'test'], loc='upper left')
# plt.savefig("loss.jpg")
model = load_model("model.hdf5")
model.summary()
scores = model.evaluate(X_train, y_train, verbose=0)
print(
'Accuracy on training data: {}% \n Error on training data: {}'.format(
scores[1], 1 - scores[1]))
scores2 = model.evaluate(X_test, y_test, verbose=0)
print('Accuracy on test data: {}% \n Error on test data: {}'.format(
scores2[1], 1 - scores2[1]))
def test_he_normal(self):
tensor_shape = (5, 6, 4, 2)
with self.cached_session():
fan_in, _ = init_ops._compute_fans(tensor_shape)
std = np.sqrt(2. / fan_in)
self._runner(init_ops.he_normal(seed=123),
tensor_shape,
target_mean=0.,
target_std=std)
def test_he_normal(self):
tensor_shape = (5, 6, 4, 2)
with self.cached_session():
fan_in, _ = init_ops._compute_fans(tensor_shape)
std = np.sqrt(2. / fan_in)
self._runner(
init_ops.he_normal(seed=123),
tensor_shape,
target_mean=0.,
target_std=std)
def build(self, fixlen_feature_columns):
self.input_dict = OrderedDict()
self.emb_matrix_dict = OrderedDict()
self.y = tf.placeholder(dtype=tf.int32, shape=[None, 1], name='y')
# 二阶部分
sparse_vec = []
for feat in fixlen_feature_columns:
if isinstance(feat, SparseFeat):
self.input_dict[feat.name] = tf.placeholder(dtype=tf.int32, shape=[None, 1], name=feat.name)
self.emb_matrix_dict[feat.name] = tf.Variable(initial_value=tf.random.normal(shape=[feat.vocabulary, feat.embedding_dim]))
vec = tf.nn.embedding_lookup(self.emb_matrix_dict[feat.name], self.input_dict[feat.name])
vec = tf.reshape(vec, shape=[-1, feat.embedding_dim])
sparse_vec.append(vec)
concat_vec =tf.concat(axis=1, values=sparse_vec)
second_ord_logits = tf.layers.dense(inputs=concat_vec, units=1, activation=tf.nn.relu, kernel_initializer=init_ops.he_normal(), kernel_regularizer=tf.nn.l2_normalize)
# 一阶部分
for feat in fixlen_feature_columns:
if isinstance(feat, DenseFeat):
self.input_dict['dense'] = tf.placeholder(dtype=tf.float32, shape=[None, feat.feature_num], name='dense')
first_ord_logits = tf.layers.dense(inputs=self.input_dict['dense'], units=1, activation=tf.nn.relu, kernel_initializer=init_ops.he_normal(), kernel_regularizer=tf.nn.l2_normalize)
logits = first_ord_logits + second_ord_logits
self.prob = tf.sigmoid(logits)