摘要: Recent breakthroughs in deep learning and artificial intelligence technologies have enabled numerous mobileapplications. While traditional computation paradigms rely on mobile sensing and cloud computing, deep learningimplemented ... 展开 Recent breakthroughs in deep learning and artificial intelligence technologies have enabled numerous mobileapplications. While traditional computation paradigms rely on mobile sensing and cloud computing, deep learningimplemented on mobile devices provides several advantages. These advantages include low communication bandwidth,small cloud computing resource cost, quick response time, and improved data privacy. Research and development ofdeep learning on mobile and embedded devices has recently attracted much attention. This paper provides a timelyreview of this fast-paced field to give the researcher, engineer, practitioner, and graduate student a quick grasp on therecent advancements of deep learning on mobile devices. In this paper, we discuss hardware architectures for mobiledeep learning, including Field Programmable Gate Arrays (FPGA), Application Specific Integrated Circuit (ASIC), andrecent mobile Graphic Processing Units (GPUs). We present Size, Weight, Area and Power (SWAP) considerations andtheir relation to algorithm optimizations, such as quantization, pruning, compression, and approximations that simplifycomputation while retaining performance accuracy. We cover existing systems and give a state-of-the-industry review ofTensorFlow, MXNet, Mobile AI Compute Engine (MACE), and Paddle-mobile deep learning platform. We discussresources for mobile deep learning practitioners, including tools, libraries, models, and performance benchmarks. Wepresent applications of various mobile sensing modalities to industries, ranging from robotics, healthcare and multimedia,biometrics to autonomous drive and defense. We address the key deep learning challenges to overcome,including low quality data, and small training/adaptation data sets. In addition, the review provides numerous citationsand links to existing code bases implementing various technologies. These resources lower the user’s barrier to entryinto the field of mobile deep learning. 收起