An artificial neural network-based strategy is developed to learn committor-consistent transition pathways, providing insight into rare events in biomolecular systems.