Work

Combining Relevance and Magnitude for Resource-Saving DNN Pruning

Year: 2025

Type: article

Abstract: Pruning neural networks, i.e., removing some of their parameters whilst retaining their accuracy, is one of the main ways to reduce the latency of a machine learning pipeline, especially in resource- and/or bandwidth-constrained scenarios. In this context, the pruningtechnique, i.e., how to choose the parameters to remove, is critical to the system performance. In this paper, we propose a novel pruning approach, called FlexRel and predicated upon combining training-time and inference-time information, namely, parameter magnitude and relevance, in order to improve the resulting accuracy whilst saving both computational resources and bandwidth. Our performance evaluation shows that FlexRel is able to achieve higher pruning factors, saving over 35% bandwidth for typical accuracy targets.

Source: IEEE Network

DOI: 10.1109/mnet.2025.3556212

Authors: Carla Fabiana Chiasserini, Francesco Malandrino, Nuria Molner, Zhao Zhiqiang

Institutions: Politecnico di Torino, Consorzio Nazionale Interuniversitario per le Telecomunicazioni, Institute of Electronics, Computer and Telecommunication Engineering, Universitat Politècnica de València

Language: English

Topic: Advanced Neural Network Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funders: European Commission, HORIZON EUROPE Framework Programme

Awards: Advancing 6G-RAN through multi-technology, multi-sensor fusion, multi-band and multi-static perception, G933377125