SWAP-NAS: Sample-Wise Activation Patterns for Ultra-fast NAS

Authors: Yameng Peng, Andy Song, Haytham M. Fayek, Vic Ciesielski, Xiaojun Chang

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Comprehensive experiments validate its robust generalisation and superior performance across five benchmark search spaces, i.e., stack-based and cell-based, and seven tasks, i.e., image classification, object detection, autoencoding and jigsaw puzzle, outperforming 15 existing training-free metrics including recent proposed NWOT and Zi Co.
Researcher Affiliation Academia Yameng Peng Andy Song Haytham M. Fayek Vic Ciesielski Xiaojun Chang ,ξ School of Computing Technologies, RMIT University, Australia University of Technology Sydney, ξMohamed bin Zayed University of Artificial Intelligence
Pseudocode Yes Algorithm 1 SWAP-NAS
Open Source Code Yes 1Our code is available at https://github.com/pym1024/SWAP.
Open Datasets Yes Image classification tasks: CIFAR-10 / CIFAR-100 (Krizhevsky, 2009), Image Net-1k (Deng et al., 2009) and Image Net16-120 (Chrabaszcz et al., 2017).
Dataset Splits Yes For example, Spearman s rank correlation coefficient between regularised SWAP-Score and CIFAR-100 validation accuracies on NAS-Bench-201 networks is 0.90, significantly higher than 0.80 from the second-best metric, NWOT. These extensive studies follow the same setup as NAS-Bench-Suite-Zero (Krishnakumar et al., 2022), which is a standardised framework for verifying the effectiveness of training-free metrics.
Hardware Specification Yes All experiments are conducted on a single Tesla V100 GPU.
Software Dependencies No The paper does not provide specific version numbers for software dependencies or libraries used in the experiments.
Experiment Setup Yes The hyper-parameters, such as batch size, input data, sampled architectures and random seeds are fixed and consistently applied to all training-free metrics as NAS-Bench-Suite-Zero. The networks training strategy and hyper-parameters are exactly following the setup in DARTS (Liu et al., 2019). SWAP-NAS-A (µ=0.9, σ=0.9), SWAP-NAS-B (µ=1.2, σ=1.2), SWAP-NAS-C (µ=1.5, σ=1.5). SWAP-NAS (µ=25, σ=25).