Edge Device
Edge devices are resource-constrained computing units performing computation closer to data sources, aiming to reduce latency, bandwidth usage, and privacy concerns associated with cloud computing. Current research focuses on optimizing deep learning models (e.g., CNNs, LLMs, GNNs) for edge deployment through techniques like model compression (quantization, pruning, knowledge distillation), efficient parallel processing (pipeline parallelism, tensor parallelism), and federated learning. This work is significant for enabling the deployment of sophisticated AI applications, such as autonomous driving and medical imaging analysis, on low-power devices, thereby expanding the accessibility and applicability of advanced technologies.
Papers
Communication-Efficient Federated Learning over Wireless Channels via Gradient Sketching
Vineet Sunil Gattani, Junshan Zhang, Gautam Dasarathy
Efficient End-to-End 6-Dof Grasp Detection Framework for Edge Devices with Hierarchical Heatmaps and Feature Propagation
Kaiqin Yang, Yixiang Dai, Guijin Wang, Siang Chen