progressive neural architecture search (pnas)

PNAS is a method to search for optimal CNN architectures, speeding up the search process by learning a surrogate model of architecture performance and using this model to choose the most promising architectures to be evaluated.