Figure 22.

Using a dilated CNN to predict the methane uptake of COFs assembled by Mercado et al.²⁹⁷ For this example, we use dilated convolutions to extract correlations from the XRPD pattern (a). We can then pass the output to some hidden layers to predict the methane uptake (b). We overfit to the training set, but can also get decent performance on the test set without major tuning of the model. MAPE is an acronym for the mean absolute percentage error.