“…In addition to the HER, ML algorithms were employed based on DFT-calculated data for the fast screening of efficient NRR and CO 2 RR electrocatalysts. 105 For instance, graph-based convolutional neural network (GCNN) was applied for the accelerated screening of SACs for the NRR. The results show superior NRR selectivity over the HER with overpotentials of 0.44 V, 0.40 V, 0.24 V, 0.60 V, 0.17 V, 0.17 V, 0.64 V, 0.37 V and 0.58 V, respectively, for SACs embedded in MBenes, defect-engineered 2D-materials, and 2D p-conjugated polymers, TaB, NbTe 2 , NbB, HfTe 2 , MoB, MnB, HfSe 2 , TaSe 2 and Nb.…”