On Optimizing Resource Allocation for MIMO-NOMA Downlink
We consider multiple-input multiple-output (MIMO) non-orthogonal multiple access (NOMA) downlink channels with zero-forcing beamforming transmission. The base station has multiple transmit antennas while all mobile devices have a single receive antenna. To increase spectral efficiency, 2 active users with highly correlated channels may be paired to share the same beamforming vector and thus, will interfere fully with each other. Superposition coding and successive interference cancellation are applied to decode messages for each user in a pair. The performance of the pair depends on the accuracy of the channel direction information (CDI) and increases with the CDI rate. With limited total transmit power and total CDI rate, we would like to optimize the transmit power and CDI rate for each user to achieve max-min fairness for all users in a cell. In this talk, we will discuss different allocation schemes that perform close to the optimum with varying complexity. Some numerical examples show that the resulting rate performance with the proposed allocation is increased by 100% over that with the uniform allocation.
Neural Network Design based on Algorithm Unrolling and Its Applications
This tutorial introduces algorithm unrolling methods to design neural networks with iterative algorithms. Algorithm unrolling methods have received a lot of attention because they provide interpretable neural network architectures that exhibit high performance. As examples applied to communication/signal processing problems, we present how to design neural networks to solve the sparse signal recovery problems and MIMO detection problems.
When Deep Unfolding Meets Control Engineering
Deep unfolding is a technique for tuning parameters for accelerating the convergence of iterative algorithms and has been actively studied in various fields including wireless communications. In this talk, we provide an overview of the speaker's work on the use of the deep unfolding technique in the field of control engineering. We specifically talk about some recent results on feedback control system design and model predictive control. Furthermore, we will review how the speaker came to know about the deep unfolding technique from the perspective of interdisciplinary research, and also try to discuss the affinity between deep unfolding and control theory research.