I believe that the best way to understand models is to reproduce the model script by hands. The goal of this blog post is to help my-past-self and someone who is stack at the similar problems in understanding Keras's RNN model. Nevertheless, there are not many good, concrete and simple explanations about the role of this parameter.
Hindsight, these questions show my lack of understanding in back propagation through time (BPTT) algorithms. Why do I need to specify the length of time series when the model is meant to handle a sequence of potentially infinite length? Where does this parameter come into play in the definition of the RNN model? Well, I was very confused with this parameter at first. batch_shape = (N of time series in a batch, the length of time series, N of features)."In theory" this may be true.īut when it comes to implementation of the RNN model in Keras, practitioners need to specify a "length of time series" in batch_shape: People say that RNN is great for modeling sequential data because it is designed to potentially remember the entire history of the time series to predict values. Recurrent Neural Network (RNN) has been successful in modeling time series data.
0 kommentar(er)
