Initial rate of H2 production (kinetics model) Final H2 concentration (yield model)   RMSE R2 RMSE R2   Base A Base B Base A Base B Base A Base B Base A Base B Autoprediction 0.003 0.001 1 1 0.15 0.001 0.8 1 k-fold cross-validation (k=10) 0.086 0.124 0.791 1 0.237 0.135 0.661 1 Validation 0.308 0.265 0.882 0.91 0.358 0.353 0.92 0.86

Table 5: Performance of ANN models built on the “Base A” and “Base B” datasets. These were composed of input and output vectors derived from respectively 32 and 12 different simulations of the H2 producing system. Shown are the Root Mean Square Errors (RMSE) and the coefficient of determination (R2) of the kinetics and yield models. Results for autoprediction and k-fold cross validation were evaluated on the learning datasets themselves. Results for validation were evaluated on the independent “Base V” dataset (see Method for details).