Communication-Efficient Split Learning Based on Analog Communication and Over the Air Aggregation
This addresses communication efficiency for privacy-preserving collaborative inference in bandwidth-limited scenarios, representing an incremental improvement over standard split learning methods.
The paper tackles the communication bottleneck in split learning for many agents with limited bandwidth by proposing a framework using analog communication and over-the-air aggregation, which maintains constant communication cost and significantly outperforms digital implementations in efficiency as agent numbers increase.
Split-learning (SL) has recently gained popularity due to its inherent privacy-preserving capabilities and ability to enable collaborative inference for devices with limited computational power. Standard SL algorithms assume an ideal underlying digital communication system and ignore the problem of scarce communication bandwidth. However, for a large number of agents, limited bandwidth resources, and time-varying communication channels, the communication bandwidth can become the bottleneck. To address this challenge, in this work, we propose a novel SL framework to solve the remote inference problem that introduces an additional layer at the agent side and constrains the choices of the weights and the biases to ensure over the air aggregation. Hence, the proposed approach maintains constant communication cost with respect to the number of agents enabling remote inference under limited bandwidth. Numerical results show that our proposed algorithm significantly outperforms the digital implementation in terms of communication-efficiency, especially as the number of agents grows large.