Communication-Efficient Edge AI: Algorithms and Systems

Abstract

Artificial intelligence (AI) has achieved remarkable breakthroughs in a wide range of fields, ranging from speech processing, image classification to drug discovery. This is driven by the explosive growth of data, advances in machine learning (especially deep learning), and easy access to vastly powerful computing resources. Particularly, the wide scale deployment of edge devices (e.g., IoT devices) generates an unprecedented scale of data, which provides the opportunity to derive accurate models and develop various intelligent applications at the network edge. However, such enormous data cannot all be sent from end devices to the cloud for processing, due to the varying channel quality, traffic congestion and/or privacy concerns. By pushing inference and training processes of AI models to edge nodes, edge AI has emerged as a promising alternative. AI at the edge requires close cooperation among edge devices, such as smart phones and smart vehicles, and edge servers at the wireless access points and base stations, which however result in heavy communication overheads. In this paper, we present a comprehensive survey of the recent developments in various techniques for overcoming these communication challenges. Specifically, we first identify key communication challenges in edge AI systems. We then introduce communication-efficient techniques, from both algorithmic and system perspectives for training and inference tasks at the network edge. Potential future research directions are also highlighted.

Federated learning for 6G communications: Challenges, methods, and future directions

“…Since the FL involves thousands of devices participating during model training, communication is a critical bottleneck for FL being widely used in 6G [11]. Previous studies [6], [8], [10], [20]- [24] has made many efforts to improve the communication efficiency of FL system. Furthermore, it is challenging for FL networks to achieve communication in the FL networks is synchronized with the local calculation of the device [8], [10], [25].…”

“…Previous studies [6], [8], [10], [20]- [24] has made many efforts to improve the communication efficiency of FL system. Furthermore, it is challenging for FL networks to achieve communication in the FL networks is synchronized with the local calculation of the device [8], [10], [25]. To make the FL model suitable for 6G networks with massive, heterogeneous devices and networks, it is necessary to develop a communication-efficient method, which can greatly reduce the number of gradients exchanged between the devices and the cloud instead of all gradients information.…”

“…1) Communication-efficient FL: System Level: From a system perspective, data distribution (e.g., non-independent and identical distribution), device distribution (e.g., heterogeneous devices across regions and networks), computation methods (e.g., decentralized and centralized), and communication mechanisms (e.g., synchronous and asynchronous scheme) have different impacts on communication efficiency in different application scenarios [8].…”

“…Although FL brings high potential for AI-empowered 6G and significantly improves privacy-sensitive applications with 6G communication, FL is still in the early stages of development and is facing new challenges in 6G scenarios. In this paper, we first introduce the core challenges of FL in 6G communications including (i) large communication cost due to multiple communication rounds for model updates and aggregation [8]; (ii) security problems caused by heterogeneous and diverse participating entities, e.g., poisoning attacks and backdoor attacks [7], [9]; (iii) privacy problems resulted from gradient leakage attacks and membership inference attacks…”

Communication-Computation Trade-Off in Resource-Constrained Edge Inference

“…Edge AI consists of edge training, i.e., to train DNN models based on data distributed at different devices, and edge inference, i.e., to provide DNN-based inference at resource-constrained devices. While communicationefficient methods for edge training have received significant attention [4], the counterpart on edge inference is less well investigated. This article aims to fill this gap and introduce new design problems and methodologies for edge inference by presenting a delicate trade-off between communication overhead and on-device computation cost.…”