Efficient Training Management for Mobile Crowd-Machine Learning: A Deep Reinforcement Learning Approach

In this letter, we consider the concept of mobile crowd-machine learning (MCML) for a federated learning model. The MCML enables mobile devices in a mobile network to collaboratively train neural network models required by a server while keeping data on the mobile devices. The MCML thus addresses da...

Full description

Saved in:
Bibliographic Details
Published in:IEEE wireless communications letters 2019-10, Vol.8 (5), p.1345-1348
Main Authors: Anh, Tran The, Luong, Nguyen Cong, Niyato, Dusit, Kim, Dong In, Wang, Li-Chun
Format: Article
Language:English
Subjects:
Algorithms
Artificial intelligence
Bandwidths
Computer simulation
Data models
Decisions
deep reinforcement learning
Electronic devices
Energy conservation
Energy consumption
federated learning
Heuristic algorithms
Machine learning
Mobile communication systems
Mobile crowd
Mobile handsets
Neural networks
Resource management
Servers
Training
Wireless networks
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this letter, we consider the concept of mobile crowd-machine learning (MCML) for a federated learning model. The MCML enables mobile devices in a mobile network to collaboratively train neural network models required by a server while keeping data on the mobile devices. The MCML thus addresses data privacy issues of traditional machine learning. However, the mobile devices are constrained by energy, CPU, and wireless bandwidth. Thus, to minimize the energy consumption, training time, and communication cost, the server needs to determine proper amounts of data and energy that the mobile devices use for training. However, under the dynamics and uncertainty of the mobile environment, it is challenging for the server to determine the optimal decisions on mobile device resource management. In this letter, we propose to adopt a deep Q -learning algorithm that allows the server to learn and find optimal decisions without any a priori knowledge of network dynamics. Simulation results show that the proposed algorithm outperforms the static algorithms in terms of energy consumption and training latency.
ISSN:2162-2337
2162-2345
DOI:10.1109/LWC.2019.2917133