Dear All,

I am interrested in how the input's temporal sturcture of a recurrent neural network (or LSTM/GRU) should be constructed? Generally we have a 3 dimensional tensor (matrix) with dimensions

(number of samples) x (number of features) x (timestep)

The (windowed) timesteps should overlap or should not? If they should, how much? Are there any general guideline?

Eg. timesteps are overlapped a possible input:

|123456|

 |234567|

  |345678|

or

|123456|

    |456789|

...

 

E.g. timesteps are not overlapped:

|123456|

       |789abc|

...

Thanks a lot,

Bálint

More Bálint Gyires-Tóth's questions See All
Similar questions and discussions