FrankenSplit: Efficient Neural Feature Compression with Shallow Variational Bottleneck Injection for Mobile Edge Computing

arXiv:2302.10681v4 Announce Type: replace-cross
Abstract: The rise of mobile AI accelerators allows latency-sensitive applications to execute lightweight Deep Neural Networks (DNNs) on the client side. However, critical applications require powerful models that edge devices cannot host and must therefore offload requests, where the high-dimensional data will compete for limited bandwidth. This work proposes shifting away from focusing on executing shallow layers of partitioned DNNs. Instead, it advocates concentrating the local resources on variational compression optimized for machine interpretability. We introduce …

abstract accelerators ai accelerators applications arxiv client compression computing cs.ai cs.dc cs.lg data devices edge edge computing edge devices eess.iv feature however latency mobile mobile ai mobile edge computing networks neural networks type will