def make_default_model(self):
self.model.add(
KERAS_LSTM(128,
input_shape=(self.input_shape[0], self.input_shape[1])))
self.model.add(Dropout(0.5))
self.model.add(Dense(32, activation='relu'))
self.model.add(Dense(16, activation='tanh'))
def make(cls,
input_shape: int,
rnn_size: int,
hidden_size: int,
dropout: float = 0.5,
n_classes: int = 6,
lr: float = 0.001):
"""
搭建模型
Args:
input_shape (int): 特征维度
rnn_size (int): LSTM 隐藏层大小
hidden_size (int): 全连接层大小
dropout (float, optional, default=0.5): dropout
n_classes (int, optional, default=6): 标签种类数量
lr (float, optional, default=0.001): 学习率
"""
model = Sequential()
model.add(KERAS_LSTM(
rnn_size,
input_shape=(1, input_shape))) # (time_steps = 1, n_feats)
model.add(Dropout(dropout))
model.add(Dense(hidden_size, activation='relu'))
# model.add(Dense(rnn_size, activation='tanh'))
model.add(Dense(n_classes, activation='softmax')) # 分类层
optimzer = Adam(lr=lr)
model.compile(loss='categorical_crossentropy',
optimizer=optimzer,
metrics=['accuracy'])
return cls(model)
def make_default_model(self):
"""
Makes the LSTM model with keras with the default hyper parameters.
"""
self.model.add(
KERAS_LSTM(128,
input_shape=(self.input_shape[0], self.input_shape[1])))
self.model.add(Dropout(0.5))
self.model.add(Dense(32, activation='relu'))
self.model.add(Dense(16, activation='tanh'))
def make_model(self):
self.model.add(KERAS_LSTM(128, input_shape=(1, self.input_shape)))
self.model.add(Dropout(0.5))
self.model.add(Dense(32, activation='relu'))
def make_model(self, rnn_size, hidden_size, dropout=0.5, **params):
self.model.add(KERAS_LSTM(
rnn_size,
input_shape=(1, self.input_shape))) # (time_steps = 1, n_feats)
self.model.add(Dropout(dropout))
self.model.add(Dense(hidden_size, activation='relu'))
