Device Deep
Device deep learning focuses on performing deep neural network training and inference directly on resource-constrained edge devices like smartphones and microcontrollers, prioritizing efficiency and privacy. Current research emphasizes developing novel algorithms like sparse backpropagation and hybrid pipeline parallelism to reduce computational costs and memory usage, alongside model compression techniques and specialized hardware accelerators (NPUs, FPGAs) for improved performance. This field is crucial for enabling privacy-preserving machine learning in applications ranging from mobile AI to IoT devices, driving advancements in both algorithm design and hardware optimization.
Papers
November 12, 2024
October 28, 2024
September 11, 2024
August 15, 2024
July 8, 2024
June 3, 2024
February 21, 2024
October 29, 2023
October 19, 2023
October 11, 2023
August 1, 2023
July 22, 2023
June 5, 2023
May 30, 2023
May 6, 2023
April 13, 2023
April 10, 2023
February 7, 2023