About this Research Topic
We refer to this emerging challenge as “Real-Time High-Performance Computing (HPC) on Edge Devices” that is essential to efficient DNN execution. To address this challenge, researchers and engineers from both academia and industry have proposed varied solutions, including but not limited to DNN model compression and neural architecture search and optimization, designing more efficient DNN execution frameworks for edge devices, and implementing more powerful hardware for edge DNN execution (e.g., various TPUs and NPUs). These solutions either focus on one kind of optimizations or leverage different co-designed optimizations (e.g., model compression-compilation co-design or software-hardware co-design). These research activities are ongoing, and more results are expected in the near future, particularly considering the recent success of ChatGPT and other large DNN models.
Within the above context, this Research Topic is particularly interested in articles that aim to address the challenge of “Real-Time High-Performance Computing (HPC) on Edge Devices” for efficient DNN execution, including but not limited to:
• Compiler-aware or execution latency-aware DNN model compression, and DNN neural architecture search or optimizations
• DNN execution performance optimization on varied edge devices with compiler, runtime, or any other HPC support
• Optimizing DNN execution performance on edge with new architecture designs
• Optimizing DNN execution performance on edge with co-designs
• DNN execution performance profiling and modeling on edge devices
• Performance optimization of emerging DNNs
• Performance optimization of DNN-based applications on edge devices
• Performance optimization of On-Device Learning or Federated Learning
• Power and/or energy optimization for DNN on edge devices
Keywords: Deep Neural Networks, Edge Devices, High-Performance Computing, Compiler, Model Optimizations
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.