Filters








77 Hits in 10.2 sec

Learning to infer: RL-based search for DNN primitive selection on Heterogeneous Embedded Systems [article]

Miguel de Prado, Nuria Pazos, Luca Benini
2018 arXiv   pre-print
finds the optimal combinations of libraries and primitives to speed up the inference of CNNs on heterogeneous embedded devices.  ...  In this work, we present QS-DNN, a fully automatic search based on Reinforcement Learning which, combined with an inference engine optimizer, efficiently explores through the design space and empirically  ...  This work is supported by the Swiss State Secretariat for Education Research and Innovation (SERI) under contract number 16.0159.  ... 
arXiv:1811.07315v1 fatcat:iyyg7uupi5huzeseb22nb7ojcy

Deep Learning based Wireless Resource Allocation with Application to Vehicular Networks [article]

Le Liang, Hao Ye, Guanding Yu, Geoffrey Ye Li
2019 arXiv   pre-print
Meanwhile, deep learning, with many success stories in various disciplines, represents a promising alternative due to its remarkable power to leverage data for problem solving.  ...  We first discuss deep learning assisted optimization for resource allocation.  ...  Then graph embedding is used to learn the vector representation of each node based on topology information. The output feature vector is then input to a DNN for link scheduling.  ... 
arXiv:1907.03289v2 fatcat:stvlo3uhqnde3icvovjsrggdwu

Automated Design Space Exploration for optimised Deployment of DNN on Arm Cortex-A CPUs [article]

Miguel de Prado, Andrew Mundy, Rabia Saeed, Maurizio Denna, Nuria Pazos, Luca Benini
2020 arXiv   pre-print
The framework relies on a Reinforcement Learning search that, combined with a deep learning inference framework, automatically explores the design space and learns an optimised solution that speeds up  ...  The spread of deep learning on embedded devices has prompted the development of numerous methods to optimise the deployment of deep neural networks (DNN).  ...  Fig. 3 : 3 Architecture of QS-DNN. Complete flow: Inference on an embedded on the left, RL-based learning on the right.  ... 
arXiv:2006.05181v2 fatcat:bfqm7genmngpxf3lzbpvh6fq3y

Woodpecker-DL: Accelerating Deep Neural Networks via Hardware-Aware Multifaceted Optimizations [article]

Yongchao Liu, Yue Jin, Yong Chen, Teng Teng, Hang Ou, Rui Zhao, Yao Zhang
2020 arXiv   pre-print
In WPK, we investigated two new automated search approaches based on genetic algorithm and reinforcement learning, respectively, to hunt the best operator code configurations targeting specific hardware  ...  than TensorRT for end-to-end model inference.  ...  Conclusion WPK is part of Woodpecker that is an efficient compiler framework for heterogeneous computing based on software-hardware co-design, and targets to accelerate deep learning applications by taking  ... 
arXiv:2008.04567v1 fatcat:toy3ovoskraavcq3ju36eg4pne

Bonseyes AI Pipeline – bringing AI to you. End-to-end integration of data, algorithms and deployment tools [article]

Miguel de Prado, Jing Su, Rabia Saeed, Lorenzo Keller, Noelia Vallez, Andrew Anderson, David Gregg, Luca Benini, Tim Llewellynn, Nabil Ouerhani, Rozenn Dahyot and, Nuria Pazos
2020 arXiv   pre-print
Next generation of embedded Information and Communication Technology (ICT) systems are collaborative systems able to perform autonomous tasks.  ...  The remarkable expansion of the embedded ICT market, together with the rise and breakthroughs of Artificial Intelligence (AI), have put the focus on the Edge as it stands as one of the keys for the next  ...  On the LPDNN side, we employ LNE coupled with an RL-based search (QS-DNN) to find an optimized solution for the deployment on the target platform.  ... 
arXiv:1901.05049v3 fatcat:dgs4zxtccne5nd4zvt4dqys44y

CLAN: Continuous Learning using Asynchronous Neuroevolution on Commodity Edge Devices [article]

Parth Mannan, Ananda Samajdar, Tushar Krishna
2020 arXiv   pre-print
With every passing day, deep learning based methods are applied to solve new problems with exceptional results. The portal to the real world is the edge.  ...  We build a prototype distributed system of Raspberry Pis communicating via WiFi running NeuroEvolutionary (NE) learning and inference.  ...  A fitness value is then assigned based on how well they performed the given task and used to select a few of the fittest members.  ... 
arXiv:2008.11881v1 fatcat:al7pmoqwy5bafpdyspckbskb44

Transferable Graph Optimizers for ML Compilers [article]

Yanqi Zhou, Sudip Roy, Amirali Abdolrashidi, Daniel Wong, Peter Ma, Qiumin Xu, Hanxiao Liu, Phitchaya Mangpo Phothilimthana, Shen Wang, Anna Goldie, Azalia Mirhoseini, James Laudon
2021 arXiv   pre-print
To address these limitations, we propose an end-to-end, transferable deep reinforcement learning method for computational graph optimization (GO), based on a scalable sequential attention mechanism over  ...  Most compilers for machine learning (ML) frameworks need to solve many correlated optimization problems to generate efficient machine code.  ...  In addition to the model parallelism primitives, GO provides a learning-based optimization that can generalize across different graphs and transfer to new tasks.  ... 
arXiv:2010.12438v2 fatcat:ju26bxgmajbgfa4wtwvgrc6k2a

Orchestrating the Development Lifecycle of Machine Learning-Based IoT Applications: A Taxonomy and Survey

Bin Qian, Jie Su, Zhenyu Wen, Devki Nandan Jha, Yinhao Li, Yu Guan, Deepak Puthal, Philip James, Renyu Yang, Albert Y. Zomaya, Omer Rana, Lizhe Wang (+2 others)
2020 ACM Computing Surveys  
Function-based selection aims to choose an appropriate concept based on their functional difference.  ...  Some of these achievements are based on the combination of DL and RL, i.e., Deep Reinforcement Learning.  ...  More details of these RL algorithms can be found in the RL methods in Appendix B, and here we focus on selecting appropriate RL algorithms based on different selection criteria.Fig. 8.  ... 
doi:10.1145/3398020 fatcat:zzgfcjxjxbhnhf53dmlo63rs3i

A Survey on Deep Reinforcement Learning for Data Processing and Analytics [article]

Qingpeng Cai, Can Cui, Yiyuan Xiong, Wei Wang, Zhongle Xie, Meihui Zhang
2022 arXiv   pre-print
Next, we discuss DRL deployment on database systems, facilitating data processing and analytics in various aspects, including data organization, scheduling, tuning, and indexing.  ...  Motivated by this trend, we provide a comprehensive review of recent works focusing on utilizing DRL to improve data processing and analytics.  ...  The DRL agent could learn to understand and solve various tasks with the right incentives. First, we introduce basic foundations and practical techniques in DRL.  ... 
arXiv:2108.04526v3 fatcat:kcusgp7jzfbf7ov5os7gwf2e6i

Multi-tenant mobile offloading systems for real-time computer vision applications

Zhou Fang, Jeng-Hau Lin, Mani B. Srivastava, Rajesh K. Gupta
2019 Proceedings of the 20th International Conference on Distributed Computing and Networking - ICDCN '19  
To overcome the drawbacks, we further propose an advanced approach based on deep reinforcement learning (RL) [102] .  ...  A simple heuristic algorithm and a more advanced algorithm based on deep reinforcement learning (RL) are presented.  ...  invoke DNN inference by calling dnn model.detect.  ... 
doi:10.1145/3288599.3288634 dblp:conf/icdcn/FangLS019 fatcat:qpib2wkm7jdnfg7k64eh3hwje4

Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis [article]

Tal Ben-Nun, Torsten Hoefler
2018 arXiv   pre-print
Based on those approaches, we extrapolate potential directions for parallelism in deep learning.  ...  We then review and model the different types of concurrency in DNNs: from the single operator, through parallelism in network inference and training, to distributed deep learning.  ...  AmoebaNets outperform all existing methods, including manually engineered DNNs and RL-based searches, with 3.8% error for ImageNet and 2.13% error for CIFAR-10 (compared to 5.29% and 3.62% on the best  ... 
arXiv:1802.09941v2 fatcat:ne2wiplln5eavjvjwf5to7nwsu

Orchestrating the Development Lifecycle of Machine Learning-Based IoT Applications: A Taxonomy and Survey [article]

Bin Qian, Jie Su, Zhenyu Wen, Devki Nandan Jha, Yinhao Li, Yu Guan, Deepak Puthal, Philip James, Renyu Yang, Albert Y. Zomaya, Omer Rana, Lizhe Wang (+2 others)
2020 arXiv   pre-print
This paper provides a comprehensive and systematic survey on the development lifecycle of ML-based IoT application.  ...  Hence, orchestrating ML pipelines that encompasses model training and implication involved in holistic development lifecycle of an IoT application often leads to complex system integration.  ...  More details of these RL algorithms can be found in the RL methods in Appendix B, and here we focus on selecting appropriate RL algorithms based on different selection criteria.Fig. 8.  ... 
arXiv:1910.05433v5 fatcat:ffvjipmylve6feuzdbav2syxfu

A Survey on Reinforcement Learning for Recommender Systems [article]

Yuanguo Lin, Yong Liu, Fan Lin, Lixin Zou, Pengcheng Wu, Wenhua Zeng, Huanhuan Chen, Chunyan Miao
2022 arXiv   pre-print
To understand the challenges and relevant solutions, there should be a reference for researchers and practitioners working on RL-based recommender systems.  ...  In particular, Reinforcement Learning (RL) based recommender systems have become an emerging research topic in recent years.  ...  For example, in DNNs-based recommender systems, users with fake profiles may be generated to promote selected items.  ... 
arXiv:2109.10665v2 fatcat:wx5ghn66hzg7faxee54jf7gspq

Wi-Fi Meets ML: A Survey on Improving IEEE 802.11 Performance with Machine Learning [article]

Szymon Szott, Katarzyna Kosek-Szott, Piotr Gawłowicz, Jorge Torres Gómez, Boris Bellalta, Anatolij Zubow, Falko Dressler
2022 arXiv   pre-print
Based on this review, we identify specific open challenges and provide general future research directions.  ...  While classical optimization approaches fail in such conditions, machine learning (ML) is able to handle complexity.  ...  learned on synthetic Retraining of features improves TL accuracy [223]* RL QL 2020 S Select subchannel new devices Apply RL for subchannel selection considering PHY data to real data Near-optimal system  ... 
arXiv:2109.04786v3 fatcat:ny55qfhsnfduzcxyve5mylpr2m

Convergence of Edge Computing and Deep Learning: A Comprehensive Survey [article]

Xiaofei Wang and Yiwen Han and Victor C.M. Leung and Dusit Niyato and Xueqiang Yan and Xu Chen
2019 arXiv   pre-print
As an important enabler broadly changing people's lives, from face recognition to ambitious smart factories and cities, developments of artificial intelligence (especially deep learning, DL) based applications  ...  promoting further discussions on the fusion of edge intelligence and intelligent edge, i.e., Edge DL.  ...  Based on such tuple, OpenEI can select a matched model for a specific edge platform based on different Edge DL capabilities in an online manner.  ... 
arXiv:1907.08349v2 fatcat:4hfqgdto4fhvlguwfjxuz3ik5q
« Previous Showing results 1 — 15 out of 77 results