Let's say I have one input feature with 10 sequences of length 25 to predict the next value. So
keras will receive an input vector of (10, 25, 1). If I use
LSTM models with
stateful=True the cell states will be remembered in between batches.
However, if I have multiple sequences in each batch, is the training equivalent to having
stateful=False and a single batch of length