We propose a deep learning based channel estimation technique for wireless energy transfer. Specifically, we develop a channel learning scheme using the deep autoencoder, which learns the channel state information (CSI) at the energy transmitter based on the harvested energy feedback from the energy receiver, in the sense of minimizing the mean square error (MSE) of the channel estimation. Numerical results demonstrate that the proposed scheme learns the CSI very well and significantly outperforms the conventional scheme in terms of the channel estimation MSE as well as the harvested energy.
24 January 2019 in IEEE Wireless Communications Letters
In this paper, we study the channel state information (CSI) feedback based on the deep autoencoder (AE) considering the feedback errors and feedback delay in the frequency division duplex (FDD) massive multiple-input multiple-output (MIMO) system. We construct the deep AE by modeling the CSI feedback process, which involves feedback transmission errors and delays. The deep AE is trained by setting the delayed version of the downlink channel as the desired output. The proposed scheme reduces the impact of the feedback errors and feedback delay. Simulation results demonstrate that the proposed scheme achieves better performance than other comparable schemes.
24 May 2021 in IEEE Transactions on Communications
Federated Learning (FL) has recently received a lot of attention for large-scale privacy-preserving machine learning. However, high communication overheads due to frequent gradient transmissions decelerate FL. To mitigate the communication overheads, two main techniques have been studied: (i) local update of weights characterizing the trade-off between communication and computation and (ii) gradient compression characterizing the trade-off between communication and precision. To the best of our knowledge, studying and balancing those two trade-offs jointly and dynamically while considering their impacts on convergence has remained unresolved even though it promises significantly faster FL. In this paper, we first formulate our problem to minimize learning error with respect to two variables: local update coefficients and sparsity budgets of gradient compression who characterize trade-offs between communication and computation/precision, respectively. We then derive an upper bound of the learning error in a given wall-clock time considering the interdependency between the two variables. Based on this theoretical analysis, we propose an enhanced FL scheme, namely Fast FL (FFL), that jointly and dynamically adjusts the two variables to minimize the learning error. We demonstrate that FFL consistently achieves higher accuracies faster than similar schemes existing in the literature.
Nonorthogonal multiple access (NOMA) and multiple-input multiple-output (MIMO) are two key enablers for 5G systems. In this article, considering the practical issue that successive interference cancellation (SIC) decoding is imperfect in the real-world NOMA system, we propose a novel scheme for the downlink of the MIMO-NOMA system based on deep learning. In this scheme, both precoding and SIC decoding of the MIMO-NOMA system are jointly optimized (or learned) in the sense of minimizing total mean square error of the users’ signals. To this end, we construct the precoder and SIC decoders using deep neural networks such that the transmitted signals intended to multiple users can be properly precoded at the transmitter based on the superposition coding technique and the received signals are accurately decodable at the users by the SIC decoding. Numerical results demonstrate the effectiveness and superior performance of the proposed scheme.
29 December 2017 in IEEE Transactions on Wireless Communications
In this paper, we study rate-energy (R-E) tradeoffs for simultaneous wireless information and power transfer (SWIPT). In the existing literature, by invoking a simplistic and ideal assumption of linear energy harvesting, the R-E tradeoff performance was analyzed only for the four SWIPT schemes: the dynamic power splitting, type-I on-off power splitting (OPS), static power splitting, and time switching. Different from such works, in this work, we consider the realistic and practical scenario of nonlinear energy harvesting. Furthermore, to characterize the R-E tradeoff with nonlinear energy harvesting, we propose a new SWIPT scheme, the generalized OPS (GOPS). As a special case of the proposed GOPS, we also investigate an additional SWIPT scheme, the type-II OPS. Through the analysis based on the realistic nonlinear models reported in the literature, we derive new theoretical results on the R-E tradeoff, which are in sharp contrast to those in the existing literature obtained with linear energy harvesting. Furthermore, we provide various useful insights into the SWIPT system with nonlinear energy harvesting.
In this letter, we propose a joint pilot design and channel estimation scheme based on the deep learning (DL) technique for multiuser multiple-input multiple output (MIMO) channels. To this end, we construct a pilot designer using two-layer neural networks (TNNs) and a channel estimator using deep neural networks (DNNs), which are jointly trained to minimize the mean square error (MSE) of channel estimation. To effectively reduce the interference among the multiple users, we also use the successive interference cancellation (SIC) technique in the channel estimation process. The numerical results demonstrate that the proposed scheme considerably outperforms the linear minimum mean square error (LMMSE) based channel estimation scheme.