My question is that let’s say I am training an autoencoder which is CNN based with same length of time series data. Then, during testing, if I feed in a much longer length data, will the model reconstruct the data well? Or the model only works well with the test sequence that has similar length as training sequence? Thanks in advance.