A.Data Annotation (Tensorboard is not an annotation tool)

B.Visualize the changes in loss and accuracy during deep learning training

C.Automatic parameter tuning (what the algorithm needs to do)

D.Visualizing high-dimensional distributions

A.LSTM is controllable in terms of the memory it retains, while GRU is not

B.There is not much difference between standard LSTM and GRU, and both are better than ordinary RNN

C.LSTM can process longer sequences than GRU

D.GRU structure is simpler than LSTM, and has one less gate than LSTM, so the calculation speed is faster than LSTM

A.False

B.True

A.Extracting features (convolutional layer)

B.Dimensionality reduction (pooling layer)

C.Classification

D.Change image size (image scaling, downsampling, upsampling)