W Wu, P Yang, W Zhang, C Zhou… - IEEE Transactions on …, 2020 - ieeexplore.ieee.org
… -intensive deep neural network (DNN) inference services, … In this article, we investigate the collaborative DNNinference … Specifically, sampling rate adaption, inference task offloading…
W Zhang, D Yang, H Peng, W Wu… - … 2020-2020 IEEE …, 2020 - ieeexplore.ieee.org
… allocation for deep neural network (DNN) inference in the … DNNinference tasks, a resource management problem is formulated with the objective of maximizing the average inference …
H Cho, P Oh, J Park, W Jung, J Lee - Proceedings of the Twenty-Fourth …, 2019 - dl.acm.org
… Deep RL platform, called FA3C. Traditionally, FPGA-based DNN accelerators have mainly focused on inference … Our platform targets both inference and training using single-precision …
C Dong, M Shafiq, MM Al Dabel… - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
… algorithm locally and get an inference scheme based on local … In this paper, we investigate DNNinference acceleration in … First, we formulate task offloading of DNNinference among …
M De Prado, N Pazos, L Benini - 2019 Design, Automation & …, 2019 - ieeexplore.ieee.org
… CNNs’ inference latency may become a bottleneck for Deep Learning adoption by … -DNN, a fully automatic search based on Reinforcement Learning which, combined with an inference …
YG Kim, CJ Wu - 2020 53rd Annual IEEE/ACM international …, 2020 - ieeexplore.ieee.org
… Therefore, AutoScale leverages a reinforcement learning … to maximize the DNNinference energy efficiency while … an in-depth characterization of DNNinference execution on mobile and …
… We focus on deep neural network (DNN) based classification … level and delay performance of DNNinference via device-edge … DNNinference in industrial IoT via deepreinforcement …
S Huang, A Ankit, P Silveira, R Antunes… - Proceedings of the 26th …, 2021 - dl.acm.org
… We also propose an automated quantization flow powered by deepreinforcement learning to search for the best quantization configuration in the large design space. Our evaluation …
… In order to mitigate the potential oscillation in the DNNinference results, we adopt the duplicate Q method from [15], which maintains two Q value estimates for each state-action pair and …