Intelligent MIMO Detection Using Meta Learning

Abstract

In a K-best detector for multiple-input-multiple-output (MIMO) systems, the value of K needs to be sufficiently large to achieve near-maximum-likelihood (ML) performance. By treating K as a variable that can be adjusted according to a fitting function of some learnable coefficients, an intelligent MIMO detection network based on deep neural networks (DNN) is proposed to reduce complexity of the detection algorithm with little performance degradation. In particular, the proposed intelligent detection algorithm uses meta learning to learn the coefficients of the fitting function for K to circumvent the problem of learning K directly. The idea of network fusion is used to combine the learning results of the meta learning component networks. Simulation results show that the proposed scheme achieves near-ML detection performance while its complexity is close to that of linear detectors. Besides, it also exhibits strong ability of fast training.

Meta-Learning for Wireless Communications: A Survey and a Comparison to GNNs

“…For example, in the problem of symbol detection, the policy was defined as the mapping from the received symbol to the transmitted symbol, which relies on the channel realization [12], [54], [61], [62]. Therefore, detecting symbols in different channel realizations was defined as different tasks in these works.…”

“…In [39], the design of an optimizer was formulated as a learning problem, where a long short-term memory (LSTM) neural network was meta-trained to optimize a common gradient for different tasks to facilitate faster convergence. This method was used in [61] and [62] for detection, where tasks were defined as detecting symbols under different channel realizations. Specifically, an expectation propagation algorithm for MIMO detection was unfolded by replacing the key factors as trainable parameters in [61], which were optimized using the gradients calibrated by a meta-trained LSTM.…”

“…Specifically, an expectation propagation algorithm for MIMO detection was unfolded by replacing the key factors as trainable parameters in [61], which were optimized using the gradients calibrated by a meta-trained LSTM. In [62], a CNN was designed to determine the number of candidates for each iteration in the K-best algorithm for symbol detection, where the gradient of the CNN during retraining was calibrated by a meta-trained LSTM.…”

Learning Statistically Robust MIMO Detection with Imperfect CSI

“…Recently, the great advancement of deep learning (DL) has motivated the researches on the DL-aided wireless communications [21]- [25]. In particular, DL has been successfully applied in the physical layer communication techniques, such as channel estimation [26], [27], MIMO precoding [28], [29], and, as our main interest, MIMO detection [30]- [35], which have been thoroughly reviewed in [24]. The DL-based designs can be categorized into data-driven and model-driven types.…”