NN-Former: Rethinking Graph Structure in Neural Architecture Representation
This work addresses neural architecture search efficiency for deep learning practitioners, though it appears incremental as it builds on existing GNN and transformer approaches.
The paper tackles the problem of predicting neural network attributes like accuracy and latency by addressing limitations in existing graph-based representation methods, proposing NN-Former which combines GNNs and transformers to better capture sibling nodes in DAG topology, achieving promising performance in both accuracy and latency prediction.
The growing use of deep learning necessitates efficient network design and deployment, making neural predictors vital for estimating attributes such as accuracy and latency. Recently, Graph Neural Networks (GNNs) and transformers have shown promising performance in representing neural architectures. However, each of both methods has its disadvantages. GNNs lack the capabilities to represent complicated features, while transformers face poor generalization when the depth of architecture grows. To mitigate the above issues, we rethink neural architecture topology and show that sibling nodes are pivotal while overlooked in previous research. We thus propose a novel predictor leveraging the strengths of GNNs and transformers to learn the enhanced topology. We introduce a novel token mixer that considers siblings, and a new channel mixer named bidirectional graph isomorphism feed-forward network. Our approach consistently achieves promising performance in both accuracy and latency prediction, providing valuable insights for learning Directed Acyclic Graph (DAG) topology. The code is available at https://github.com/XuRuihan/NNFormer.