Fig. 8From: Deep convolutional architectures for extrapolative forecasts in time-dependent flow problemsResidual Block (left) and architecture of CNN and TCN (right). For TCN, the input vector \(V \ \in \) \({\mathbb {R}}^{n_s \times n_t}\) and \(c = n_{s}\). For CNN, the input vector \(V \ \in \) \({\mathbb {R}}^{n_t \times n_s}\) and \(c = 1\)Back to article page