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 learn- ing pipeline, especially in resource- and/or bandwidth-constrained scenarios. In this context, the pruning tech- nique, i.e., how to choose the parameters to remove, is crit- ical to the system performance. In this paper, we propose a novel pruning approach, called FlexRel and predicated upon combining training-time and inference-time infor- mation, 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: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.15274433

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

Institution: Politecnico di Torino

Language: English

Topic: Advanced Neural Network Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W6911640515

FWCI:

Cites:

Cited by:

Related to: