Fig. 3From: Deep convolutional architectures for extrapolative forecasts in time-dependent flow problemsAutoencoder architectureBack to article page