我从Keras文档知道,我们可以使用tensorboard回调来可视化图层,例如:
from keras.callbacks import Tensorboard
log_dir = ...
X_train = ...
X_test = ...
tensorboard = TensorBoard(batch_size=batch_size,
embeddings_freq=1,
embeddings_layer_names=['features'],
embeddings_data=x_test)
这将显示使用X_test数据嵌入名为features
的图层。
我想知道是否有可能获得两个不同数据集(即X_train和X_test)的嵌入可视化效果,例如:
tensorboard = TensorBoard(batch_size=batch_size,
embeddings_freq=1,
embeddings_layer_names=['features'],
embeddings_data=[X_train, x_test])