def get_autoencoder_feature(data,
max_features,
max_len,
model,
tokenizer=None):
'''
@description: 获取autoencoder 特征
@param {type}
train, train data set
test, test data set
max_features, max_features
max_len, max_len
model, autoencoder model
tokenizer, autoencoder tokenizer
@return: DataFrame of train and test
'''
# 格式化数据
X, _ = format_data(data,
max_features,
max_len,
tokenizer=tokenizer,
shuffle=True)
# 使用autoencoder 的encoder 进行预测
data_ae = pd.DataFrame(model.predict(X, batch_size=64,
verbose=1).max(axis=1),
columns=['ae' + str(i) for i in range(max_len)])
return data_ae
def get_autoencoder_feature(data,
max_features,
max_len,
model,
tokenizer=None):
'''
@description: get_autoencoder_feature
@param {type}
train, train data set
test, test data set
max_features, max_features
max_len, max_len
model, autoencoder model
tokenizer, autoencoder tokenizer
@return: DataFrame of train and test
'''
### TODO
# 1. 返回autoencoder embedding
x, _ = format_data(data,
max_features,
max_len,
tokenizer=tokenizer,
shuffle=True)
data_ae = pd.DataFrame(model.predict(x, batch_size=64, verbose=1),
columns=['ae' + str(i) for i in range(max_len)])
return data_ae
def train(self, data, epochs=1):
'''
@description: Train autoencoder model
@param {type}
data, train data
epochs, train how many times
@return:
'''
# 处理数据
self.X, self.tokenizer = format_data(data,
self.max_features,
self.max_len,
shuffle=True)
self.model.fit(self.X,
self.X,
epochs=epochs,
batch_size=128,
verbose=1)
def train(self, data, epochs=10):
self.x, self.tokenizer = format_data(data, self.embedding_dim, self.max_len, shuffle=True)
self.model.fit(self.x, self.x, epochs=epochs, batch_size=32)
def get_autoencoder_feature(data, max_features, max_len, model, tokenizer=None):
x, _ = format_data(data, max_features, max_len, tokenizer=tokenizer, shuffle=True)
data_ae = pd.DataFrame(model.predict(X, batch_size=64).max(axis=1), columns=['ae' + str(i) for i in range(max_len)])
return data_ae