I\'m testing out different implementation of LSTM autoencoder on anomaly detection on 2D input. My question is not about the code itself but about understanding the underlyi