Closed jblewdv closed 5 years ago
You should probably be using a recurrent network with 4 inputs. You should pass your 10 steps of data in a loop ignoring the output until the last input where you will consider the output of the network as your final result.
I'm curious if when using a FeedforwardNetwork the inputs have to be in a 1D array? I've attempted passing 2D arrays with no success, so I wanted to ask here to make sure. If I were to pass the following as input:
x = [ [1, 2, 3, 4], [2, 3, 4, 5], [3, 4, 5, 6], .....]
Could the net recognize their to be 3 inputs (for the 3 individual arrays) instead of 4 inputs (4 coming from the length of each array)? I'm working with time series data that has 4 values at every time step and I'd like to pass in the last 10 time steps of data as input, therefore amounting to 40 inputs. Is this possible? Thanks!