Paper
1 August 2022 Pruning threshold search algorithm combined with PDARTS
Jiaxiang Ma
Author Affiliations +
Proceedings Volume 12257, 4th International Conference on Information Science, Electrical, and Automation Engineering (ISEAE 2022); 122571G (2022) https://doi.org/10.1117/12.2640465
Event: 4th International Conference on Information Science, Electrical, and Automation Engineering (ISEAE 2022), 2022, Guangzhou, China
Abstract
With the continuous improvement of network accuracy, the depth and width of the network also increase, so the requirements for storage and operation equipment are also higher and higher, which limits the application scenarios of the network. To solve this problem, many network acceleration methods have been proposed, among which thresholdbased pruning algorithm and differentiation-based network architecture search algorithm are two commonly used methods. The optimal pruning threshold is found by searching through a differentiation-based network architecture search algorithm using an incremental approach. Experiments were carried out on CIFAR-10/100 using VGG16, ResNet-56 and ResNet-18 respectively. The experimental results show that the algorithm in this paper can help the network to maintain better classification accuracy after pruning with a higher compression rate.
© (2022) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Jiaxiang Ma "Pruning threshold search algorithm combined with PDARTS", Proc. SPIE 12257, 4th International Conference on Information Science, Electrical, and Automation Engineering (ISEAE 2022), 122571G (1 August 2022); https://doi.org/10.1117/12.2640465
Advertisement
Advertisement
RIGHTS & PERMISSIONS
Get copyright permission  Get copyright permission on Copyright Marketplace
KEYWORDS
Network architectures

Optimization (mathematics)

Algorithm development

Neural networks

Computer science

Back to Top