Neural Architecture Retrieval
This addresses the challenge for researchers in comparing and connecting neural architecture designs, though it is incremental as it builds on existing graph representation techniques.
The paper tackles the problem of efficiently discovering similar neural architectures among many existing designs by defining Neural Architecture Retrieval, and proposes a method using motif-based graph division and multi-level contrastive learning, achieving superior results on evaluations with 12k real-world architectures.
With the increasing number of new neural architecture designs and substantial existing neural architectures, it becomes difficult for the researchers to situate their contributions compared with existing neural architectures or establish the connections between their designs and other relevant ones. To discover similar neural architectures in an efficient and automatic manner, we define a new problem Neural Architecture Retrieval which retrieves a set of existing neural architectures which have similar designs to the query neural architecture. Existing graph pre-training strategies cannot address the computational graph in neural architectures due to the graph size and motifs. To fulfill this potential, we propose to divide the graph into motifs which are used to rebuild the macro graph to tackle these issues, and introduce multi-level contrastive learning to achieve accurate graph representation learning. Extensive evaluations on both human-designed and synthesized neural architectures demonstrate the superiority of our algorithm. Such a dataset which contains 12k real-world network architectures, as well as their embedding, is built for neural architecture retrieval.