Antenna selection for multiple-input multiple-output systems based on deep convolutional neural networks

Abstract

Antenna selection in Multiple-Input Multiple-Output (MIMO) systems has attracted increasing attention due to the challenge of keeping a balance between communication performance and computational complexity. Recently, deep learning based methods have achieved promising performance in many application fields. This paper proposed a deep learning (DL) based antenna selection technique. First, we generated the label of training antenna systems by maximizing the channel capacity. Then, we adopted the deep convolutional neural network (CNN) on the channel matrices to explicitly exploit the massive latent cues of attenuation coefficients. Finally, we used the adopted CNN to assign the class label and then select the optimal antenna subset. Experimental results demonstrate that our method can achieve better performance than the state-of-the-art baselines for data-driven based antenna selection.

A Novel Machine Learning Aided Antenna Selection Scheme for MIMO Internet of Things

“…256 neurons and the dropout ratio of ratio = 0.2 were considered in the full connection layer. We considered (8, 8; 2, 8) and (32, 32; 2, 32) MIMO IoT systems and included the LeNet-based AS of [22] and NBAS of [7] as for performance comparison, where we employed same training data and training method to train LeNet model.…”

“…LeNet model is a state-of-the-art data-driven AS method and was shown to be capable of operating in real time scenario [22]. As a comparison, the overall online prediction time complexity of LeNet model may be expressed as 1 3 •…”

“…Recently, an increasing number of works started to focus on applying machine learning (ML) technologies in a wide range of communication applications due to the fact that ML is capable of transfering the conventional mathematical optimization problems into data-driven problems for achieving lower online realtime computational complexity [20][21][22]. More specifically, it has been well recognized that the ML schemes, i.e., convolutional neural network (CNN), exhibit excellent performance in image processing, especially in classification problems [23][24][25][26].…”

“…Additionally, ref. [22] proposed the LeNet model for receiving AS schemes, which is a single-label multi-class classification CNN model. The proposed method used convolutional structure to extract the rich features from the channel matrices, and numerical experimental results showed that the LeNet model-based AS outperformed the state-of-the-art baselines.…”

“…The binary bit of value 1 indicates that the corresponding antenna is selected, while value 0 means not selected. By contrast, the general single-label used in ML scheme is a vector composed of multiple binary bits [22] with the length of the vector equal to the total number of AS combinations of the ergodic search process, with bit value 1 corresponding to an antenna combination, noting that in single-label scheme, only one binary bit in the vector equals to 1. Table 1 shows a comparison example of our multi-label and the general single-label corresponding to N T = 4 and L t = 2, where it may be seen that as the number of AS combinations increases, the multi-label scheme may achieve significantly lower complexity than that of the conventional single-label ML schemes.…”