In this paper, we propose an orthogonal frequency-division multiplexing system supported by the compressed sensing assisted index modulation, termed as (OFDM-CSIM), applied to millimeter-wave (mmWave) communications. In the OFDM-CSIM mmWave system, information is conveyed not only by the classic constellation symbols but also by the on/off status of subcarriers, where the size of constellation symbols and the number of active subcarriers can be beneficially configured for maximizing the system's throughput. We conceive a machine learning (ML) assisted adaptive OFDM-CSIM mmWave system, which simultaneously benefits from the OFDM with index modulation (IM), compressed sensing (CS) and the hybrid beamforming techniques. Specifically, a ML-assisted link adaptation scheme is designed based on the k-nearest neighbors (k-NN) algorithm with the objective to maximize the system's throughput. Our studies show that the proposed ML-assisted link adaptation is capable of providing higher throughput than the conventional threshold-based link adaptation when different antenna structures are considered. Furthermore, the achievable data rates of four types of antenna arrays, including uniform linear array (ULA), uniform rectangular planar array (URPA), uniform circle planar array (UCPA) and uniform cylindrical array (UCYA), are investigated and compared over mmWave channels. The simulation results show that the UCYA achieves the highest data rate among these antenna arrays.

The efficiency of link adaptation in wireless communications relies greatly on the accuracy of channel knowledge and transmission mode selection. In this paper, a novel deep learning based link adaptation framework is proposed for the orthogonal frequency-division multiplexing (OFDM) systems with compressed-sensing-assisted index modulation, termed as OFDM-CSIM, communicating over millimeter-wave (mmWave) channels. To achieve link adaptation, a novel multi-layer sparse Bayesian learning (SBL) algorithm is proposed for accurately and instantaneously providing the required channel state information. Meanwhile, a deep neural networks (DNN)-assisted adaptive modulation algorithm is proposed to choose the best possible transmission mode to maximize the achievable throughput. Simulation results show that the proposed multi-layer SBL algorithm enables more accurate channel estimation than the conventional techniques. The DNN-based adaptive modulator is capable of achieving a higher throughput than the learning-assisted solution based on the k nearest neighbor (k-NN) algorithm, and also the classic average signal-to-noise ratio (SNR)-based solutions. Moreover, analysis shows that both the multi-layer SBL algorithm and the DNN-assisted adaptive modulator achieve better performance than their respective conventional counterparts while at a significantly lower computational complexity cost.