FrankenSplit: Efficient Neural Feature Compression with Shallow Variational Bottleneck Injection for Mobile Edge Computing
This addresses bandwidth constraints for mobile edge computing applications requiring powerful models, though it appears incremental as it builds on existing compression and offloading approaches.
The paper tackles the problem of offloading high-dimensional data from edge devices to servers by proposing a variational compression method optimized for machine interpretability, achieving 60% lower bitrate than a state-of-the-art method without accuracy loss and up to 16x faster than existing codec standards.
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 a novel framework for resource-conscious compression models and extensively evaluate our method in an environment reflecting the asymmetric resource distribution between edge devices and servers. Our method achieves 60% lower bitrate than a state-of-the-art SC method without decreasing accuracy and is up to 16x faster than offloading with existing codec standards.