Signal Prediction using sequence models

Can we train a encoder decoder sequence model to presdict the future of a time series signal. I want to predict the noise of a given signal.


That’s different than predicting the expected value.

Actually we have a nuclear signals where the baseline is fluctuating. So if we can create a model that can predict the fluctuating baseline then we may be able to find what the actual value of the signal was at the moment it reaches the peak. As the peak value is distorted by the signal value.

Sounds like an interesting situation to model.