DeepAI AI Chat
Log In Sign Up

Network Slimming by Slimmable Networks: Towards One-Shot Architecture Search for Channel Numbers

03/27/2019
by   Jiahui Yu, et al.
2

We study how to set channel numbers 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 solution, 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. Notably, by setting optimized channel numbers, our AutoSlim-MobileNet-v2 at 305M FLOPs achieves 74.2 (301M FLOPs), and even 0.2 AutoSlim-ResNet-50 at 570M FLOPs, without depthwise convolutions, achieves 1.3 better accuracy than MobileNet-v1 (569M FLOPs). Code and models will be available at: https://github.com/JiahuiYu/slimmable_networks

READ FULL TEXT
12/20/2019

AtomNAS: Fine-Grained End-to-End Neural Architecture Search

Designing of search space is a critical problem for neural architecture ...
10/15/2021

Joint Channel and Weight Pruning for Model Acceleration on Moblie Devices

For practical deep neural network design on mobile devices, it is essent...
12/07/2022

Slimmable Pruned Neural Networks

Slimmable Neural Networks (S-Net) is a novel network which enabled to se...
01/13/2023

Adaptive Neural Networks Using Residual Fitting

Current methods for estimating the required neural-network size for a gi...
05/20/2019

DARC: Differentiable ARchitecture Compression

In many learning situations, resources at inference time are significant...
12/21/2018

Slimmable Neural Networks

We present a simple and general method to train a single neural network ...
07/13/2021

Exploring Channel Probing to Determine Coherent Optical Transponder Configurations in a Long-Haul Network

We use channel probing to determine the best transponder configurations ...

Code Repositories

slimmable_networks

Slimmable Networks, AutoSlim, and Beyond, ICLR 2019, and ICCV 2019


view repo