Adaptive Compression-Aware Split Learning and Inference for Enhanced Network Efficiency

TL;DR

The paper tackles network efficiency for edge-cloud AI by enhancing split learning with an adaptive compression-decompression module. It introduces deprune to dramatically reduce training-bandwidth by progressively tightening the communication budget and prune to enable rapid model generation at multiple budgets via transfer learning. Together, these methods improve training efficiency (up to about 6x faster) and reduce network usage (up to ~30x) while preserving or improving accuracy compared with compression-aware baselines. The approach is validated on realistic testbeds and standard vision datasets, highlighting practical benefits for resource-constrained devices in edge-cloud deployments.

Abstract

The growing number of AI-driven applications in mobile devices has led to solutions that integrate deep learning models with the available edge-cloud resources. Due to multiple benefits such as reduction in on-device energy consumption, improved latency, improved network usage, and certain privacy improvements, split learning, where deep learning models are split away from the mobile device and computed in a distributed manner, has become an extensively explored topic. Incorporating compression-aware methods (where learning adapts to compression level of the communicated data) has made split learning even more advantageous. This method could even offer a viable alternative to traditional methods, such as federated learning techniques. In this work, we develop an adaptive compression-aware split learning method ('deprune') to improve and train deep learning models so that they are much more network-efficient, which would make them ideal to deploy in weaker devices with the help of edge-cloud resources. This method is also extended ('prune') to very quickly train deep learning models through a transfer learning approach, which trades off little accuracy for much more network-efficient inference abilities. We show that the 'deprune' method can reduce network usage by 4x when compared with a split-learning approach (that does not use our method) without loss of accuracy, while also improving accuracy over compression-aware split-learning by 4 percent. Lastly, we show that the 'prune' method can reduce the training time for certain models by up to 6x without affecting the accuracy when compared against a compression-aware split-learning approach.

Figures (16)

• Figure 1: Illustration of different weaker devices opting for nearby computation sources, for AI tasks
• Figure 2: System Overview: split learning representation across multiple devices with the compression-decompression module involved
• Figure 3: Illustration of Compression and Decompression modules
• Figure 4: Data traffic in the Wi-Fi channel during the benchmark execution.
• Figure 5: Throughput for each data source during the benchmark execution. The plots are interpolated linearly to make them more readable.