from keras.layers.core import Dense, Dropout, Activation
from keras.layers.embeddings import Embedding
import keras
import numpy as np
x_train = np.load(args["train_data"] + "/train/seq/" + "RNN_data_train_" + args["len_seq"] + "_" + str(j) + ".npy")
y_train = np.load(args["train_data"] + "/train/label/" + "RNN_label_train_" + args["len_seq"] + "_" + str(j) + ".npy")
# 将整型的类别标签转为onehot编码
y_in = keras.utils.to_categorical(y_train, 2)
# len_s = int(args["len_seq"])
print("train_data:", x_train)
print("train_data_shape:", x_train.shape)
print("train_label:", y_in)
print("train_label_shape:", y_in.shape)
model = Sequential()
model.add(Embedding(22, 16, input_length=11))
model.add(Dropout(0.2))
model.add(LSTM(16, return_sequences=True, activation='relu'))
model.add(LSTM(16, return_sequences=False))
model.add(Dropout(0.2))
model.add(Dense(10, activation='relu'))
model.add(Dropout(0.2))
model.add(Dense(2, activation='softmax'))
model.summary()
callback = keras.callbacks.EarlyStopping(monitor='val_loss', patience=30)
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.RMSprop(lr=0.001),
metrics=['accuracy'])
train_history = model.fit(x_train, y_in, batch_size=8000, epochs=400,
verbose=2, validation_split=0.2, shuffle=True,callbacks=[callback])
model.save(modelfile/rnn_mode_new_dropout.h5）
添加层数，神经元个数试试