Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective

Abstract

Neural Architecture Search (NAS) has been explosively studied to automate the discovery of top-performer neural networks. Current works require heavy training of supernet or intensive architecture evaluations, thus suffering from heavy resource consumption and often incurring search bias due to truncated training or approximations. Can we select the best neural architectures without involving any training and eliminate a drastic portion of the search cost? We provide an affirmative answer, by proposing a novel framework called training-free neural architecture search (TE-NAS). TE-NAS ranks architectures by analyzing the spectrum of the neural tangent kernel (NTK) and the number of linear regions in the input space. Both are motivated by recent theory advances in deep networks and can be computed without any training and any label. We show that: (1) these two measurements imply the trainability and expressivity of a neural network; (2) they strongly correlate with the network's test accuracy. Further on, we design a pruning-based NAS mechanism to achieve a more flexible and superior trade-off between the trainability and expressivity during the search. In NAS-Bench-201 and DARTS search spaces, TE-NAS completes high-quality search but only costs 0.5 and 4 GPU hours with one 1080Ti on CIFAR-10 and ImageNet, respectively. We hope our work inspires more attempts in bridging the theoretical findings of deep networks and practical impacts in real NAS applications. Code is available at: https://github.com/VITA-Group/TENAS.

arXiv commentaccepted as ICLR 2021 poster

References in corpus (21)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Neural Architecture Search with Reinforcement Learning
• DARTS: Differentiable Architecture Search
• Neural Tangent Kernel: Convergence and Generalization in Neural Networks
• Neural Architecture Optimization
• PC-DARTS: Partial Channel Connections for Memory-Efficient Architecture Search
• A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets
• Searching for Efficient Multi-Scale Architectures for Dense Image Prediction
• SNAS: Stochastic Neural Architecture Search
• DARTS+: Improved Differentiable Architecture Search with Early Stopping
• On Lazy Training in Differentiable Programming
• Single Path One-Shot Neural Architecture Search with Uniform Sampling
• Deep Neural Networks with Random Gaussian Weights: A Universal Classification Strategy?
• NATS-Bench: Benchmarking NAS Algorithms for Architecture Topology and Size
• Scaling Limits of Wide Neural Networks with Weight Sharing: Gaussian Process Behavior, Gradient Independence, and Neural Tangent Kernel Derivation
• Picking Winning Tickets Before Training by Preserving Gradient Flow
• Deep ReLU Networks Have Surprisingly Few Activation Patterns
• Semi-Supervised Neural Architecture Search
• Neural Architecture Search without Training
• How to Train Your Super-Net: An Analysis of Training Heuristics in Weight-Sharing NAS
• HourNAS: Extremely Fast Neural Architecture Search Through an Hourglass Lens

Cited by in corpus (21)

• Sustainable AI: Environmental Implications, Challenges and Opportunities
• When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations
• Neural Architecture Search without Training
• Evaluating Efficient Performance Estimators of Neural Architectures
• One Proxy Device Is Enough for Hardware-Aware Neural Architecture Search
• Dataset Distillation with Infinitely Wide Convolutional Networks
• Evolutionary Spiking Neural Networks: A Survey
• Zen-NAS: A Zero-Shot NAS for High-Performance Deep Image Recognition
• NASI: Label- and Data-agnostic Neural Architecture Search at Initialization
• Understanding and Accelerating Neural Architecture Search with Training-Free and Theory-Grounded Metrics
• On the Privacy Risks of Cell-Based NAS Architectures
• Generic Neural Architecture Search via Regression
• ProxyBO: Accelerating Neural Architecture Search via Bayesian Optimization with Zero-cost Proxies
• BaLeNAS: Differentiable Architecture Search via the Bayesian Learning Rule
• FLASH: Fast Neural Architecture Search with Hardware Optimization
• Deep Learning Theory through the Lens of Statistical Physics: A Topical Review
• On the Security Risks of AutoML
• Mutually-aware Sub-Graphs Differentiable Architecture Search
• Training BatchNorm Only in Neural Architecture Search and Beyond
• A Neural Tangent Kernel Perspective of Infinite Tree Ensembles
• Connection Sensitivity Matters for Training-free DARTS: From Architecture-Level Scoring to Operation-Level Sensitivity Analysis