Machine Learning and Deep Learning Methods for Wireless Network Applications
Abel C. H. Chen,Wen-Kang Jia,Feng-Jang Hwang,Genggeng Liu,Fangying Song,Lianrong Pu
DOI: https://doi.org/10.1186/s13638-022-02196-2
2022-01-01
EURASIP Journal on Wireless Communications and Networking
Abstract:Wireless networks have been widely adopted and introduced in the areas of engineering, manufacturing, weather monitoring, transportation, etc., to collect data to improve the quality of decision making, but issues arise, such as large volumes of data, incomplete and incompatible data sets, and noise data that prevent from realizing the true value and exploiting their full potentials.Machine learning and deep learning methods have been used as powerful tools to perform feature detection/extraction and trend estimation/forecasting in wireless networks applications.Supervised machine learning methods, for example, neural network (NN) [1, 2], convolutional neural network (CNN) [3], and recurrent neural network (RNN) [4], can be exploited in the applications pertinent to prediction and classification, whereas unsupervised machine learning methods such as restricted Boltzmann machine (RBM), deep belief network (DBN), deep Boltzmann machine (DBM), auto-encoder (AE), and denoising auto-encoder (DAE) [5] can be utilized for data denoising and model generalization.Furthermore, the reinforcement learning methods, including generative adversarial networks (GANs) and deep Q-networks (DQNs) [6], are tools for generative networks and discriminative networks to optimize the contesting process in a zero-sum game framework.These well-developed methods can contribute substantially, to better improve predictions and classifications in the relevant applications, but there are some issues and limitations that require further attention from the research communities.Topics covered in this issue are categorized into the following six themes: (1) networking optimization, (2) unmanned aerial vehicle (UAV) applications, (3) wireless sensor networks, (4) network security, (5) mobile positioning, and (6) image-based applications.This special issue has collected a total of 24 published papers from the USA, Australia, China, India, Pakistan, South Korea, and Vietnam to ensure that high-quality papers with significant results are selected and published.The accepted papers categorized into the aforementioned eight themes are briefly introduced below. Networking optimizationFour papers on networking optimization are listed as follows: (1) "Decentralized computation offloading for multi-user mobile edge computing: a deep reinforcement learning approach, " by Zhao Chen and Xiaodong Wang from China and the USA [7]; (2)