NeurIPS 2020

BRP-NAS: Prediction-based NAS using GCNs

Meta Review

The paper analysis of existing latency predictors is solid and the proposed method seems to provide a real improvement to the task of predicting latency. The overall performance lift is not demonstrated in a fully convincing manner and there could be some room for more extensive experiments. Still, all things considered, a solid method for predicting latency is relevant to NAS making this paper interesting to the NeurIPS community