Combining Relevance and Magnitude for Resource-Aware DNN Pruning
This work addresses efficiency challenges for deploying deep learning models in bandwidth-limited environments, representing an incremental improvement over existing pruning techniques.
The paper tackles the problem of pruning neural networks to reduce latency and bandwidth in resource-constrained scenarios by proposing FlexRel, a method that combines parameter magnitude and relevance, achieving over 35% bandwidth savings while maintaining accuracy.
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 pruning technique, 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.