Python 将模型与Keras合并'；函数API
我是深入学习的新手，正在自学一些NLP概念。我试图通过学习Quora的句子相似性评分模型来理解这一点，正如本优秀教程中所解释的 这段代码是很久以前的，在尝试让它在新数据集上运行时，我发现模型合并API已经被弃用了。我已经坐了很长一段时间，努力合并模型（因为我不太了解复杂性） 有人能帮我确认一下我是否用正确的方法转换了这个吗？我的主要问题是输入序列已填充到Python 将模型与Keras合并'；函数API,python,merge,keras,keras-layer,Python,Merge,Keras,Keras Layer,我是深入学习的新手，正在自学一些NLP概念。我试图通过学习Quora的句子相似性评分模型来理解这一点，正如本优秀教程中所解释的 这段代码是很久以前的，在尝试让它在新数据集上运行时，我发现模型合并API已经被弃用了。我已经坐了很长一段时间，努力合并模型（因为我不太了解复杂性） 有人能帮我确认一下我是否用正确的方法转换了这个吗？我的主要问题是输入序列已填充到max_len的40，那么这是否意味着我必须指定一个形状为40的输入层 站点上的旧代码： tk = text.Tokenizer(nb_word
max_lentk = text.Tokenizer(nb_words=200000)
max_len = 40
tk.fit_on_texts(list(data.question1.values) + list(data.question2.values.astype(str)))
x1 = tk.texts_to_sequences(data.question1.values)
x1 = sequence.pad_sequences(x1, maxlen=max_len)
x2 = tk.texts_to_sequences(data.question2.values.astype(str))
x2 = sequence.pad_sequences(x2, maxlen=max_len)
word_index = tk.word_index
ytrain_enc = np_utils.to_categorical(y)
embeddings_index = {}
f = open('data/glove.840B.300d.txt')
for line in tqdm(f):
values = line.split()
word = values[0]
coefs = np.asarray(values[1:], dtype='float32')
embeddings_index[word] = coefs
f.close()
print('Found %s word vectors.' % len(embeddings_index))
embedding_matrix = np.zeros((len(word_index) + 1, 300))
for word, i in tqdm(word_index.items()):
embedding_vector = embeddings_index.get(word)
if embedding_vector is not None:
embedding_matrix[i] = embedding_vector
max_features = 200000
filter_length = 5
nb_filter = 64
pool_length = 4
model = Sequential()
print('Build model...')
model1 = Sequential()
model1.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model1.add(TimeDistributed(Dense(300, activation='relu')))
model1.add(Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,)))
model2 = Sequential()
model2.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model2.add(TimeDistributed(Dense(300, activation='relu')))
model2.add(Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,)))
model3 = Sequential()
model3.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model3.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model3.add(Dropout(0.2))
model3.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model3.add(GlobalMaxPooling1D())
model3.add(Dropout(0.2))
model3.add(Dense(300))
model3.add(Dropout(0.2))
model3.add(BatchNormalization())
model4 = Sequential()
model4.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model4.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model4.add(Dropout(0.2))
model4.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model4.add(GlobalMaxPooling1D())
model4.add(Dropout(0.2))
model4.add(Dense(300))
model4.add(Dropout(0.2))
model4.add(BatchNormalization())
model5 = Sequential()
model5.add(Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2))
model5.add(LSTM(300, dropout_W=0.2, dropout_U=0.2))
model6 = Sequential()
model6.add(Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2))
model6.add(LSTM(300, dropout_W=0.2, dropout_U=0.2))
merged_model = Sequential()
merged_model.add(Merge([model1, model2, model3, model4, model5, model6], mode='concat'))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(1))
merged_model.add(Activation('sigmoid'))
tk = text.Tokenizer(nb_words=200000, filters='!"#$%&*+,/;<>?@\^`{|}~')
max_len = 40
tk.fit_on_texts(list(data.log1.values) + list(data.log2.values.astype(str)))
x1 = tk.texts_to_sequences(data.log1.values)
x1 = sequence.pad_sequences(x1, maxlen=max_len)
x2 = tk.texts_to_sequences(data.log2.values.astype(str))
x2 = sequence.pad_sequences(x2, maxlen=max_len)
word_index = tk.word_index
ytrain_enc = np_utils.to_categorical(y)
embeddings_index = {}
f = open('data/glove.840B.300d.txt')
for line in tqdm(f):
values = line.split()
word = values[0]
coefs = np.asarray(values[1:], dtype='float32')
embeddings_index[word] = coefs
f.close()
print('Found %s word vectors.' % len(embeddings_index))
embedding_matrix = np.zeros((len(word_index) + 1, 300))
for word, i in tqdm(word_index.items()):
embedding_vector = embeddings_index.get(word)
if embedding_vector is not None:
embedding_matrix[i] = embedding_vector
max_features = 200000
filter_length = 5
nb_filter = 64
pool_length = 4
input_layer = Input(shape=(40,))
print('Build model...')
# Converting everything to Keras' functional API
# We don't need model1 = Sequential() since the functional API specifies this
# implicitly
model1 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model1 = TimeDistributed(Dense(300, activation='relu'))(model1)
model1 = Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,))(model1)
model2 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model2 = TimeDistributed(Dense(300, activation='relu'))(model2)
model2 = Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,))(model2)
model3 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model3 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length,
border_mode='valid', activation='relu',
subsample_length=1)(model3)
model3 = Dropout(0.2)(model3)
model3 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length,
border_mode='valid', activation='relu',
subsample_length=1)(model3)
model3 = GlobalMaxPooling1D()(model3)
model3 = Dropout(0.2)(model3)
model3 = Dense(300)(model3)
model3 = Dropout(0.2)(model3)
model3 = BatchNormalization()(model3)
model4 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model4 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='valid',
activation='relu', subsample_length=1)(model4)
model4 = Dropout(0.2)(model4)
model4 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='valid',
activation='relu', subsample_length=1)(model4)
model4 = GlobalMaxPooling1D()(model4)
model4 = Dropout(0.2)(model4)
model4 = Dense(300)(model4)
model4 = Dropout(0.2)(model4)
model4 = BatchNormalization()(model4)
model5 = Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2)(input_layer)
model5 = LSTM(300, dropout_W=0.2, dropout_U=0.2)(model5)
model6 = Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2)(input_layer)
model6 = LSTM(300, dropout_W=0.2, dropout_U=0.2)(model6)
merged_model = concatenate([model1, model2, model3, model4, model5, model6])
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
# merged_model = Dense(1)(merged_model)
# merged_model = Activation('sigmoid')(merged_model)
predictions = Dense(1, activation='sigmoid')(merged_model)
model = Model(inputs=input_layer, outputs=predictions)
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
model.summary()
checkpoint = ModelCheckpoint('weights.h5', monitor='val_acc', save_best_only=True, verbose=2)
model.fit([x1, x2, x1, x2, x1, x2], y=y, batch_size=384, nb_epoch=200,
verbose=1, validation_split=0.1, shuffle=True, callbacks=[checkpoint])
tk = text.Tokenizer(nb_words=200000)
max_len = 40
tk.fit_on_texts(list(data.question1.values) + list(data.question2.values.astype(str)))
x1 = tk.texts_to_sequences(data.question1.values)
x1 = sequence.pad_sequences(x1, maxlen=max_len)
x2 = tk.texts_to_sequences(data.question2.values.astype(str))
x2 = sequence.pad_sequences(x2, maxlen=max_len)
word_index = tk.word_index
ytrain_enc = np_utils.to_categorical(y)
embeddings_index = {}
f = open('data/glove.840B.300d.txt')
for line in tqdm(f):
values = line.split()
word = values[0]
coefs = np.asarray(values[1:], dtype='float32')
embeddings_index[word] = coefs
f.close()
print('Found %s word vectors.' % len(embeddings_index))
embedding_matrix = np.zeros((len(word_index) + 1, 300))
for word, i in tqdm(word_index.items()):
embedding_vector = embeddings_index.get(word)
if embedding_vector is not None:
embedding_matrix[i] = embedding_vector
max_features = 200000
filter_length = 5
nb_filter = 64
pool_length = 4
model = Sequential()
print('Build model...')
model1 = Sequential()
model1.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model1.add(TimeDistributed(Dense(300, activation='relu')))
model1.add(Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,)))
model2 = Sequential()
model2.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model2.add(TimeDistributed(Dense(300, activation='relu')))
model2.add(Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,)))
model3 = Sequential()
model3.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model3.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model3.add(Dropout(0.2))
model3.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model3.add(GlobalMaxPooling1D())
model3.add(Dropout(0.2))
model3.add(Dense(300))
model3.add(Dropout(0.2))
model3.add(BatchNormalization())
model4 = Sequential()
model4.add(Embedding(len(word_index) + 1,
300,
weights=[embedding_matrix],
input_length=40,
trainable=False))
model4.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model4.add(Dropout(0.2))
model4.add(Convolution1D(nb_filter=nb_filter,
filter_length=filter_length,
border_mode='valid',
activation='relu',
subsample_length=1))
model4.add(GlobalMaxPooling1D())
model4.add(Dropout(0.2))
model4.add(Dense(300))
model4.add(Dropout(0.2))
model4.add(BatchNormalization())
model5 = Sequential()
model5.add(Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2))
model5.add(LSTM(300, dropout_W=0.2, dropout_U=0.2))
model6 = Sequential()
model6.add(Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2))
model6.add(LSTM(300, dropout_W=0.2, dropout_U=0.2))
merged_model = Sequential()
merged_model.add(Merge([model1, model2, model3, model4, model5, model6], mode='concat'))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(300))
merged_model.add(PReLU())
merged_model.add(Dropout(0.2))
merged_model.add(BatchNormalization())
merged_model.add(Dense(1))
merged_model.add(Activation('sigmoid'))
tk = text.Tokenizer(nb_words=200000, filters='!"#$%&*+,/;<>?@\^`{|}~')
max_len = 40
tk.fit_on_texts(list(data.log1.values) + list(data.log2.values.astype(str)))
x1 = tk.texts_to_sequences(data.log1.values)
x1 = sequence.pad_sequences(x1, maxlen=max_len)
x2 = tk.texts_to_sequences(data.log2.values.astype(str))
x2 = sequence.pad_sequences(x2, maxlen=max_len)
word_index = tk.word_index
ytrain_enc = np_utils.to_categorical(y)
embeddings_index = {}
f = open('data/glove.840B.300d.txt')
for line in tqdm(f):
values = line.split()
word = values[0]
coefs = np.asarray(values[1:], dtype='float32')
embeddings_index[word] = coefs
f.close()
print('Found %s word vectors.' % len(embeddings_index))
embedding_matrix = np.zeros((len(word_index) + 1, 300))
for word, i in tqdm(word_index.items()):
embedding_vector = embeddings_index.get(word)
if embedding_vector is not None:
embedding_matrix[i] = embedding_vector
max_features = 200000
filter_length = 5
nb_filter = 64
pool_length = 4
input_layer = Input(shape=(40,))
print('Build model...')
# Converting everything to Keras' functional API
# We don't need model1 = Sequential() since the functional API specifies this
# implicitly
model1 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model1 = TimeDistributed(Dense(300, activation='relu'))(model1)
model1 = Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,))(model1)
model2 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model2 = TimeDistributed(Dense(300, activation='relu'))(model2)
model2 = Lambda(lambda x: K.sum(x, axis=1), output_shape=(300,))(model2)
model3 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model3 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length,
border_mode='valid', activation='relu',
subsample_length=1)(model3)
model3 = Dropout(0.2)(model3)
model3 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length,
border_mode='valid', activation='relu',
subsample_length=1)(model3)
model3 = GlobalMaxPooling1D()(model3)
model3 = Dropout(0.2)(model3)
model3 = Dense(300)(model3)
model3 = Dropout(0.2)(model3)
model3 = BatchNormalization()(model3)
model4 = Embedding(len(word_index) + 1, 300, weights=[embedding_matrix],
input_length=40, trainable=False)(input_layer)
model4 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='valid',
activation='relu', subsample_length=1)(model4)
model4 = Dropout(0.2)(model4)
model4 = Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='valid',
activation='relu', subsample_length=1)(model4)
model4 = GlobalMaxPooling1D()(model4)
model4 = Dropout(0.2)(model4)
model4 = Dense(300)(model4)
model4 = Dropout(0.2)(model4)
model4 = BatchNormalization()(model4)
model5 = Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2)(input_layer)
model5 = LSTM(300, dropout_W=0.2, dropout_U=0.2)(model5)
model6 = Embedding(len(word_index) + 1, 300, input_length=40, dropout=0.2)(input_layer)
model6 = LSTM(300, dropout_W=0.2, dropout_U=0.2)(model6)
merged_model = concatenate([model1, model2, model3, model4, model5, model6])
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
merged_model = Dense(300)(merged_model)
merged_model = PReLU()(merged_model)
merged_model = Dropout(0.2)(merged_model)
merged_model = BatchNormalization()(merged_model)
# merged_model = Dense(1)(merged_model)
# merged_model = Activation('sigmoid')(merged_model)
predictions = Dense(1, activation='sigmoid')(merged_model)
model = Model(inputs=input_layer, outputs=predictions)
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
model.summary()
checkpoint = ModelCheckpoint('weights.h5', monitor='val_acc', save_best_only=True, verbose=2)
model.fit([x1, x2, x1, x2, x1, x2], y=y, batch_size=384, nb_epoch=200,
verbose=1, validation_split=0.1, shuffle=True, callbacks=[checkpoint])
tk=text.Tokenizer（nb#u words=200000，filters='！“#$%&*+，/；？@^`{124;}~）
最大长度=40
文本上的tk.fit（list（data.log1.values）+list（data.log2.values.astype（str）））
x1=tk.文本到序列（data.log1.值）
x1=序列。焊盘序列（x1，最大长度=最大长度）
x2=tk.text_到_序列（data.log2.values.astype（str））
x2=序列。焊盘序列（x2，最大长度=最大长度）
word\u index=tk.word\u index
ytrain_enc=np_utils.to_category（y）
嵌入_索引={}
f=打开（'data/globet.840B.300d.txt'））
对于tqdm（f）中的行：
values=line.split（）
字=值[0]
coefs=np.asarray（值[1:]，dtype='float32'）
嵌入索引[word]=coefs
f、 关闭（）
打印（'找到%s个字向量。'%len（嵌入索引））
嵌入矩阵=np.0（（len（单词索引）+1300））
对于word，TQM中的i（word_index.items（））：
嵌入向量=嵌入索引.get（word）
如果嵌入_向量不是无：
嵌入矩阵[i]=嵌入向量
最大功能=200000
过滤器长度=5
nb_过滤器=64
池长=4
输入层=输入（形状=（40，））
打印（'生成模型…'）
#将所有内容转换为Keras的功能API
#我们不需要model1=Sequential（），因为函数API指定了这一点
#含蓄地
模型1=嵌入（len（单词索引）+1300，权重=[嵌入矩阵]，
输入长度=40，可训练=False）（输入层）
model1=时间分布（密集（300，activation='relu'））（model1）
模型1=λ（λx:K.sum（x，轴=1），输出形状=（300，）（模型1）
模型2=嵌入（len（单词索引）+1300，权重=[嵌入矩阵]，
输入长度=40，可训练=False）（输入层）
模型2=时间分布（密集型（300，激活=relu'））（模型2）
模型2=λ（λx:K.sum（x，轴=1），输出_形=（300，）（模型2）
模型3=嵌入（len（单词索引）+1300，权重=[嵌入矩阵]，
输入长度=40，可训练=False）（输入层）
模型3=卷积1D（nb_过滤器=nb_过滤器，过滤器长度=过滤器长度，
border_mode='valid'，activation='relu'，
子样本长度=1）（模型3）
模型3=辍学（0.2）（模型3）
模型3=卷积1D（nb_过滤器=nb_过滤器，过滤器长度=过滤器长度，
border_mode='valid'，activation='relu'，
子样本长度=1）（模型3）
model3=globalMapooling1d（）（model3）
模型3=辍学（0.2）（模型3）
模型3=密度（300）（模型3）
模型3=辍学（0.2）（模型3）
model3=BatchNormalization（）（model3）
模型4=嵌入（len（单词索引）+1300，权重=[嵌入矩阵]，
输入长度=40，可训练=False）（输入层）
model4=卷积1d（nb_filter=nb_filter，filter_length=filter_length，border_mode='valid'，
激活='relu'，子样本长度=1）（型号4）
模型4=辍学（0.2）（模型4）
model4=卷积1d（nb_filter=nb_filter，filter_length=filter_length，border_mode='valid'，
激活='relu'，子样本长度=1）（型号4）
model4=globalMapooling1d（）（model4）
模型4=辍学（0.2）（模型4）
模型4=密度（300）（模型4）
模型4=辍学（0.2）（模型4）
model4=BatchNormalization（）（model4）
模型5=嵌入（len（单词索引）+1300，输入长度=40，辍学=0.2）（输入层）
模型5=LSTM（300，辍学率W=0.2，辍学率W=0.2）（模型5）
模型6=嵌入（len（单词索引）+1300，输入长度=40，辍学=0.2）（输入层）
模型6=LSTM（300，辍学率=0.2，辍学率=0.2）（模型6）
合并的模型=连接（[model1、model2、model3、model4、model5、model6]）
合并的_模型=批处理规范化（）（合并的_模型）
合并的_模型=密集（300）（合并的_模型）
合并的_模型=PReLU（）（合并的_模型）
合并的_模型=退出（0.2）（合并的_模型）
合并的_模型=批处理规范化（）（合并的_模型）
合并的_模型=密集（300）（合并的_模型）
合并的_模型=PReLU（）（合并的_模型）
合并的_模型=退出（0.2）（合并的_模型）
合并的_模型=批处理规范化（）（合并的_模型）
合并的_模型=密集（300）（合并的_模型）
合并的_模型=PReLU（）（合并的_模型）
合并的_模型=退出（0.2）（合并的_模型）
合并的_模型=批处理规范化（）（合并的_模型）
合并的_模型=密集（300）（合并的_模型）
合并的_模型=PReLU（）（合并的_模型）
合并的_模型=退出（0.2）（合并的_模型）
合并的_模型=批处理规范化（）（合并的_模型）
合并的_模型=密集（300）（合并的_模型）
合并的_模型=PReLU（）（合并的_模型）
合并的_模型=退出（0.2）（合并的_模型）
合并的_模型=批处理规范化（）（合并的_模型）
#合并的_模型=密集（1）（合并的_模型）
#合并模型=激活（'sigmoid'）（合并模型）
预测=密集（1，激活='sigmoid'）（合并模型）
模型=模型（输入=输入层，输出=预测）
compile（loss='binary\u crossentropy'，optimizer='adam'，metrics=['accurity']）
model.summary（）
检查点=模型检查点（'weights.h5'，monitor='val\u acc'，save\u best\u only=True，verbose=2）
模型拟合（[x1，x2，x1，x2，x1，x2]，y=y，批次尺寸=384，nb=200，
verbose=1，validation\u split=0.1，shuffle=True，callbacks=[checkpoint]）