The input to the neural net is wrong

[tastypenguinbacon]

X_train is basically the whole input series, so the practically, what the LSTM has to do is to select a column from the input data and undo the normalise_zero_base which does not normalize the column but does the row (which is probably wrong) and in the case of y data is definitely wrong, because you divide the all the elements but the first window_len ones by all the elements except the last window_len and you do it one by one, so basically: y_train[0] = train_data[target_col][window_len] / train_data[target_col][0] y_train[1] = train_data[target_col][window_len + 1] / train_data[target_col][1] y_train[2] = train_data[target_col][window_len + 2] / train_data[target_col][2] ...

[danipolo]

any news on that?