AutoSlim: Towards One-Shot Architecture Search for Channel NumbersDownload PDF

25 Sept 2019 (modified: 03 Jun 2023)ICLR 2020 Conference Blind SubmissionReaders: Everyone
Abstract: We study how to set the number of channels in a neural network to achieve better accuracy under constrained resources (e.g., FLOPs, latency, memory footprint or model size). A simple and one-shot approach, named AutoSlim, is presented. Instead of training many network samples and searching with reinforcement learning, we train a single slimmable network to approximate the network accuracy of different channel configurations. We then iteratively evaluate the trained slimmable model and greedily slim the layer with minimal accuracy drop. By this single pass, we can obtain the optimized channel configurations under different resource constraints. We present experiments with MobileNet v1, MobileNet v2, ResNet-50 and RL-searched MNasNet on ImageNet classification. We show significant improvements over their default channel configurations. We also achieve better accuracy than recent channel pruning methods and neural architecture search methods with 100X lower search cost. Notably, by setting optimized channel numbers, our AutoSlim-MobileNet-v2 at 305M FLOPs achieves 74.2% top-1 accuracy, 2.4% better than default MobileNet-v2 (301M FLOPs), and even 0.2% better than RL-searched MNasNet (317M FLOPs). Our AutoSlim-ResNet-50 at 570M FLOPs, without depthwise convolutions, achieves 1.3% better accuracy than MobileNet-v1 (569M FLOPs).
Keywords: AutoSlim, Neural Architecture Search, Efficient Networks, Network Pruning
TL;DR: We present an automated approach to search the number of channels in a neural network to achieve better accuracy under constrained resources (e.g., FLOPs, latency, memory footprint or model size).
Code: [![github](/images/github_icon.svg) JiahuiYu/slimmable_networks](https://github.com/JiahuiYu/slimmable_networks) + [![Papers with Code](/images/pwc_icon.svg) 5 community implementations](https://paperswithcode.com/paper/?openreview=H1gz_nNYDS)
Data: [CIFAR-10](https://paperswithcode.com/dataset/cifar-10), [ImageNet](https://paperswithcode.com/dataset/imagenet)
Original Pdf: pdf
7 Replies

Loading