Future wireless networks are anticipated to evolve by aerial communication platforms. Nonetheless, the operational lifespan and efficacy of transceivers such as unmanned aerial vehicle (UAVs) and Internet of Things (IoT) devices are strictly prohibited by their constrained onboard power sources. This paper focuses on an aerial network configuration where a UAV harvests solar power to serve energy-limited IoT devices through simultaneous wireless information and power transfer. In this setup, the UAV and the IoT devices, each are equipped with energy and data buffers. This system also benefits from rate splitting multiple access for efficient interference management. Upon optimizing the system efficacy, we formulate a long-term resource allocation problem to maximize the time-averaged energy efficiency. To address this stochastic and non-convex optimization problem, we propose a multi-stage solution strategy. Firstly, by leveraging Lyapunov optimization theory, we transform the long-term stochastic problem into an equivalent deterministic short-term form. Next, by recasting this equivalent problem into Markov decision process, we propose a resource allocation mechanism based on actor-critic hindsight experience replay (AC-HER), tailored to capture the problem dynamics and optimize its variables. Moreover, given the UAV high mobility and the system reconfigurations, we fortify the trained AC-HER model with meta-learning strategy, enhancing its adaptability to system variations. Simulations verified that the proposed resource allocation strategy considerably outperforms its counterparts.

This study investigates the application of non-orthogonal multiple access (NOMA) to enable massive connectivity for a stacked intelligent metasurface (SIM) system that performs signal processing in the electromagnetic wave domain. To realize the full potential of NOMA assisted SIM systems, a radio resource allocation problem is accordingly formulated to jointly optimize the key decision variables, including the decoding order of users, the transmit power at the base station, as well as the phase shift at the SIM. By adhereing to the users’ quality-of-service (QoS) requirements, as well as the power budget of the base station, the problem is aimed at maximizing the admission rate of the system. Due to the problem’s inherent non-convexity and complexity, we recast it as a Markov decision process and employ a quantile regression deep Q-network (QRDQN) agent to optimize the decision variables. Recognizing the mobility of users and the dynamic reconfiguration of the system, we further enhance the QR-DQN model’s adaptability and generalization capabilities by incorporating a meta-learning strategy. Simulation results demonstrate that integrating NOMA with SIM systems yields a significant increase of 39% in the average number of served users compared to the conventional orthogonal multiple access based approach. The proposed resource allocation mechanism additionally improves deep deterministic policy gradient (DDPG) in literature by 27% in the number of served users.

Stacked intelligent metasurface (SIM) as an advanced signal processing paradigm enables real-time processing of electromagnetic waves at the speed of light. Benefiting from this technology, the current paper studies the downlink transmission of a wireless network, where a SIM-deployed base station (BS) serves two disjoint sets of energy- and information-oriented terminals via simultaneous wireless information and power transfer (SWIPT). Toward optimizing the performance of this system, a resource allocation problem is formulated for characterizing the fundamental trade-off between the aggregate information rate and the overall harvested energy. By virtue of its tightly-coupled and non-convex nature, we equivalently transform this problem to a Markov decision process (MDP) form. Next, we train an asynchronous advantage actor critic (A3C) agent on the MDP-reformulated problem for optimizing the transmit power of the BS and the electromagnetic response of the SIM, in a joint fashion. Subsequently, by taking into account the mobility of terminals, we further enrich the adaptability of the trained A 3 C agent via meta-learning. It is numerically revealed that incorporating SIM leads to an approximate 30 % enhancement in the energy efficiency of existing SWIPT systems.

This paper centers around a multi-antenna unmanned aerial vehicle (UAV) which leverages stacked intelligent metasurface (SIM) as a green technology for signal processing at the electromagnetic wave domain. To assess the performance of this system, a radio resource allocation problem is accordingly formulated for jointly optimizing the motion trajectory and transmit power at the UAV, as well as the electromagnetic response at the SIM as decision variables. Since the problem is non-convex and challenging to solve, we reformulate it in Markov decision process form and train a distributed distributional deep deterministic policy gradient (D4PG) agent to optimize its decision variables. Concerning the significant mobility of the UAV and thus remarkable rearrangement of the system, we enhance the adaptability and generalization of the trained D4PG model by integrating meta-learning strategy. According to simulations, wave-domain beamforming via SIM at UAV leads to 40% and 22% reduction in average energy consumption, compared fullydigital and beamspace beamforming, respectively.

This study focuses on downlink transmissions of a low earth orbit (LEO) satellite, assisted by a beyond diagonal reconfigurable intelligent surface (BD-RIS) to serve ground terminals. Toward optimizing the performance of this system, we formulate the minimization of the average age of information (AoI) achieved at ground terminals. Our formulation respects the power budget of the LEO satellite and guarantees the quality-of-service of ground terminals by optimizing the downlink transmit power at the LEO satellite and reflection coefficients at the BD-RIS as decision variables. Owing to its non-convex and tightly-coupled nature, we reformulate the problem as a Markov decision process which effectively captures its dynamics. Next, a Q-learning propagation (Q-Prop) agent is trained to optimize the decision variables. In light of the mobility of ground terminals as well as LEO satellite, this communication system is highly dynamic. Therefore, we enhance the trained Q-Prop model with meta-learning strategy, which augments its adaptability and generalization to system variances. Numerical results indicate that, in comparison to RIS-lacking and RIS-assisted counterparts, our optimised solution achieves 38% and 26% lower average AoI, respectively.

This paper centers around a multiple-input-multiple-output (MIMO) visible light communication (VLC) system, where an unmanned aerial vehicle (UAV) benefits from a light emitting diode (LED) array to serve photo-diode (PD)equipped users for illumination and communication simultaneously. Concerning the battery limitation of the UAV and considerable energy consumption of the LED array, a hybrid dimming control scheme is devised at the UAV that effectively controls the number of glared LEDs and thereby mitigates the overall energy consumption. To assess the performance of this system, a radio resource allocation problem is accordingly formulated for jointly optimizing the motion trajectory, transmit beamforming and LED selection at the UAV, assuming that channel state information (CSI) is partially available. By reformulating the optimization problem in Markov decision process (MDP) form, we propose a soft actor-critic (SAC) mechanism that captures the dynamics of the problem and optimizes its parameters. Additionally, regarding the high mobility of the UAV and thus remarkable rearrangement of the system, we enhance the trained SAC model by integrating a meta-learning strategy that enables more adaptation to system variations. By defining energy efficiency as a trade-off between the data rate and power consumption, simulations verify that upgrading a single-LED UAV by an array of 10 LEDs, exhibits 47 % and 34 % improvements in data rate and energy efficiency, albeit at the expense of 8 % more power consumption.

Unmanned aerial vehicles (UAVs) with multiple antennas have recently been explored to improve capacity in wireless networks. However, their strict energy constraint for simultaneously flying and communication tasks renders the exploration of energy-efficient multi-antenna techniques indispensable. Meanwhile, lens antenna subarrays (LASs) emerge as a promising energy-efficient multi-antenna structure that have not been previously harnessed for this purpose. In this paper, we propose a LAS-aided UAV to serve ground users in downlink transmission. We formulate a resource allocation problem aimed at initiating a trade-off between aggregate data rate of ground users and the power consumption of the UAV (energy efficiency) by optimizing the lens-based beamforming and flight trajectory of the UAV. To address this non-convex problem, we recast it in Markov decision process that captures its dynamic features and provides a framework to train an actor-critic agent. This agent is fine-tuned via hindsight experience replay for enhanced stabilization. As well, given the frequent mobility of the UAV, we fortify the trained agent with a meta-learning strategy, enhancing its adaptability to system variations. Numerically, more than 20% energy efficiency gain is achieved by incorporating a 4lens LAS for UAV, compared to its single-lens architecture in literature. Simulations also demonstrate that the proposed resource allocation strategy achieves significant superiority over counterparts in literature.

This endeavor centers around downlink transmission of a base station (BS), outfitted with a stacked intelligent metasurface (SIM) that realizes an energy-efficient wave-domain multi-user beamforming. The underlying network includes mobile devices with multimedia service requirements, including web surfing, HTTP live video streaming and voice-over-LTE (VoLTE). Rather than conventional quality-of-service (QoS) metrics, we assess the satisfaction level of users relying on quality-of-experience (QoE) criteria. Invoking mean opinion score (MOS) as the subjective measurement of QoE, we also evaluate the overall efficacy of this system by posing a resource allocation optimization problem aimed at maximizing the achievable MOS of all users, while adhering to their minimum MOS requirements and the maximum transmit power budget of the BS. Due to the interdependency of optimization variables and non-convex nature of the problem, we first reformulate it in Markov decision process, which captures its dynamic traits. Relying on the MDP model, a conservative Q-learning (CQL) agent is trained for jointly designing the optimization variables, including the BS downlink transmit power, as well as the electromagnetic response of the SIM. In light of real-time mobility of users and the resulting non-trivial network dynamism, we further utilize meta-learning technique to enhance the adaptability and generalization of the CQL agent. Numerically, it is demonstrated that, respectively, 31%, 44% and 26% superior average MOS is achieved, for web, video and audio services, compared to traditional QoS-driven resource allocation.

This research investigates a wireless system in which a base station (BS), outfitted with a stacked intelligent meta-surface (SIM), performs wave-domain multi-user beamforming in downlink transmission. The communication benefits from the assistance of an active reconfigurable intelligent surface (RIS) that amplifies incoming signal strength to extend the network coverage. System performance is evaluated through the formulation of a resource allocation optimization problem aimed at maximizing the number of served users while adhering to their quality-of-service demands and the maximum transmit power budget of the BS. Due to the complex interdependencies among optimization variables and the non-convex nature of the problem, we first reformulate it in Markov decision process, which captures its dynamic traits. Subsequently, a maximum a posteriori policy optimization (MPO) agent is trained for jointly designing the optimization variables, including the BS transmit power, the electromagnetic response of the SIM, as well as the amplitude/phase of the active RIS. In light of real-time mobility of users and non-trivial network dynamism, we invoke the integration of meta-learning technique to enhance the adaptability and generalization of the MPO model. Numerically, it is demonstrated that incorporating an active RIS upscales the number of served users by 57% and 113%, on average, in comparison with existing passive RIS-assisted and conventional SIM-enabled systems, respectively.

Reconfigurable intelligent surface (RIS) has emerged as a groundbreaking paradigm in shaping the future of wireless systems in late years. Derived from RISs, stacked intelligent metasurface (SIM), by enabling the real-time modulation of electromagnetic waves at the speed of light, and simultaneous transmitting and reflecting RIS (STAR-RIS), by drastically enhancing the coverage extension of cellular networks, appear to revolutionize the future of wireless communication. This letter delves into the synergization of SIM and STAR-RIS in a wireless system, where a base station (BS), outfitted with a SIM, benefits from a STAR-RIS to serve downlink receivers. The performance of the system is examined through the formulation of a resource allocation optimization problem, which seeks to maximize the system’s data rate, subject to constraints on receivers’ quality-of-service and the BS’s transmit power budget. Given the interdependency of variables and its inherently non-convex nature, the problem is firstly transformed into a Markov decision process form, which encapsulates its dynamic characteristics. Thereafter, a natural actor critic (NAC) agent is employed to holistically optimize the problem variables, including the transmit power at the BS, the electromagnetic response at the SIM and the reflection coefficients at the STAR-RIS. Furthermore, to account for the real-time mobility of receivers and thus the dynamism of the network, we enhance the adaptability of the trained NAC model via meta-learning technique. Simulation results reveal that the introduction of a STAR-RIS improves the system data rate, by 21% and 38% on average, compared to existing RIS-enabled and conventional SIM-based systems, respectively.

Composed of an array of low-cost radiating elements, flexible intelligent metasurfaces (FIMs) can adaptively morph their surface shapes by adjusting the positions of elements along the direction perpendicular to the surface. This adaptive morphing, not only enhances wireless channel conditions, but also significantly curtails power consumption. This paper conducts an adaptive performance analysis of a wireless system, in which a FIM-equipped base station (BS) leverages the presence of a reconfigurable intelligent surface (RIS) to facilitate downlink transmission. To rigorously evaluate the system’s efficiency, we formulate an optimization problem centered on resource allocation, with the primary objective of maximizing the network achievable data rate. This maximization is subject to multiple constraints, especially on stringent quality-of-service (QoS) requirements of users and the BS finite power budget. Due to the highly intricate interdependencies among optimization variables and the inherent non-convexity of the problem, we strategically reformulate it as a Markov decision process (MDP), encapsulating its dynamic characteristics. To derive an optimal solution, we train a deep deterministic policy gradient (DDPG) agent, relying on MDP, which simultaneously optimizes the decision variables: the BS transmit beamforming, the morphology of the FIM and the reflection coefficient matrix of the RIS. Furthermore, to accommodate the real-world challenges imposed by user mobility, we enhance the generalization of the DDPG model through the integration of meta-learning technique, thereby significantly improving its adaptability to system variations. Numerical evaluations affirm that incorporating an RIS yields a pronounced improvement in achievable network data rate, particularly when the BS transmit power budget is maintained within a moderate operational range.

Optical wireless communication (OWC) systems with multiple light-emitting diodes (LEDs) have recently been benefited from the assistance of optical reflecting intelligent surface (ORIS) to support energy-limited devices via simultaneous lightweight information and power transfer (SLIPT). This article studies the application of rate splitting multiple access (RSMA) for effective interference management and enhancing the data rate of these systems. Regarding the considerable bandwidth of the OWC band and also considerable energy consumption of the multi-LED transmitter, we formulate an energy efficiency (EE) maximization problem to jointly optimize the system variables, including transmit beamforming, LED selection, rate adaptation and ORIS element association, while adhering to the system requirements. Accordingly, we propose a dynamic resource allocation mechanism, leveraging proximal policy optimization (PPO) to accommodate system dynamism and optimize its variables. Concerning the frequent obstruction of OWC Line-of-Sight (LoS) links and consequently swift system reconfiguration, we improve the adaptability and predictability of the PPO agent by integrating Meta-learning technique. Simulations reveal that the proposed Meta-PPO algorithm has superior performance compared to the PPO method in the presence of ORIS with 76% gain. Furthermore, employing an ORIS in the proposed system model improves the performance by 51% compared to a scenario without ORIS.

Downlink transmission of a nonorthogonal visible light communication (VLC) system, empowered by autonomous aerial vehicles (AAVs), is studied for coexisting enhanced mobile broadband (eMBB), ultrareliable low-latency communication (URLLC), and massive machine-type communication (mMTC) services. A joint resource allocation problem involving user association, transmit power, and flight trajectory of AAVs is formulated, with the goal of characterizing a multiobjective tradeoff as a weighted sum of the power consumption of each AAV and the perceived Quality of Experience (QoE) of its associated eMBB users, while ensuring the service-specific requirements for eMBB, mMTC, and URLLC are met. Assuming the imperfection of channel state information (CSI), we invoke a generalized Benders decomposition (GBD) methodology, leveraging tools from convex optimization and multiagent deep reinforcement learning to address this problem. We further analytically derive the upper and lower bounds on the reward function for each AAV as a learning agent. Extensive simulations confirm that our proposed method outperforms the single-agent counterpart in the literature, with up to a 22% reduction in power consumption and a 13% gain in perceived QoE. Additionally, compared to the globally optimal brute-force method for AAV-user association, our proposed method experiences only a trivial performance loss in a small-scale scenario.

In this paper, we consider the downlink transmission of a multi-antenna base station (BS) supported by an active simultaneously transmitting and reconfigurable intelligent surface (STAR-RIS) to serve single-antenna users via simultaneous wireless information and power transfer (SWIPT). In this context, we formulate an energy efficiency maximisation problem that jointly optimises the gain, element selection and phase shift matrices of the active STAR-RIS, the transmit beamforming of the BS and the power splitting ratio of the users. With respect to the highly coupled and non-convex form of this problem, an alternating optimisation solution approach is proposed, using tools from convex optimisation and reinforcement learning. Specifically, semi-definite relaxation (SDR), difference of convex functions (DC), and fractional programming techniques are employed to transform the non-convex optimisation problem into a convex form for optimising the BS beamforming vector and the power splitting ratio of the SWIPT. Then, by integrating meta-learning with the modified deep deterministic policy gradient (DDPG) and soft actor-critical (SAC) methods, a combinatorial reinforcement learning network is developed to optimise the element selection, gain and phase shift matrices of the active STAR-RIS. Our simulations show the effectiveness of the proposed resource allocation scheme. Furthermore, our proposed active STAR-RIS-based SWIPT system outperforms its passive counterpart by 57% on average.

Optical wireless communication (OWC) systems with multiple light-emitting diodes (LEDs) have recently been explored to support energy-limited devices via simultaneous lightwave information and power transfer (SLIPT). The energy consumption, however, becomes considerable by increasing the number of incorporated LEDs. This paper proposes a joint dimming (JD) scheme that lowers the consumed power of a SLIPT-enabled OWC system by controlling the number of active LEDs. We further enhance the data rate of this system by utilizing rate splitting multiple access (RSMA). More specifically, we formulate a data rate maximization problem to optimize the beamforming design, LED selection and RSMA rate adaptation that guarantees the power budget of the OWC transmitter, as well as the quality-of-service (QoS) and an energy harvesting level for users. We propose a dynamic resource allocation solution based on proximal policy optimization (PPO) reinforcement learning. In simulations, the optimal dimming level is determined to initiate a trade-off between the data rate and power consumption. It is also verified that RSMA significantly improves the data rate.

Mounting a reconfigurable intelligent surface (RIS) on an unmanned aerial vehicle (UAV) holds promise for improving traditional terrestrial network performance. Unlike conventional methods deploying passive RIS on UAVs, this study delves into the efficacy of an aerial active RIS (AARIS). Specifically, the downlink transmission of an AARIS network is investigated, where the base station (BS) leverages rate-splitting multiple access (RSMA) for effective interference management and benefits from the support of an AARIS for jointly amplifying and reflecting the BS’s transmit signals. Considering both the non-trivial energy consumption of the active RIS and the limited energy storage of the UAV, we propose an innovative element selection strategy for optimizing the on/off status of active RIS elements, which adaptively and remarkably manages the system’s power consumption. To this end, a resource management problem is formulated, aiming to maximize the system energy efficiency (EE) by jointly optimizing the transmit beamforming at the BS, the element activation, the phase shift and the amplification factor at the active RIS, the RSMA common data rate at users, as well as the UAV’s trajectory. Due to the dynamicity nature of UAV and user mobility, a deep reinforcement learning (DRL) algorithm is designed for resource allocation, utilizing meta-learning to adaptively handle fast time-varying system dynamics. According to simulations, integrating meta-learning yields a notable 36% increase in system EE. Additionally, substituting AARIS for fixed terrestrial active RIS results in a 26% EE enhancement.

Downlink transmission of non-orthogonal multiple access visible light communication systems empowered by an unmanned aerial vehicle (UAV) is considered for multiplexing enhanced mobile broadband (eMBB) and massive machine type communication (mMTC) services. Accordingly, a resource allocation problem of joint transmit power control and motion trajectory design of the DAVs is formulated, whose goal is to characterize a multi-objective trade-off as a weighted sum of the UAVs' power consumption and the perceived quality-of-experience (QoE) of eMBB users, while ensuring the eMBB and mMTC service-specific requirements. We leverage an alternative decomposition and tools from convex optimization and actorcritic multi-agent deep reinforcement learning to address this problem in an iterative fashion. We analytically derive the upper-and lower-bounds on the reward of the DAVs as the learning agents and demonstrate that the proposed resource allocation method outperforms the similar scheme in literature, by up to 17% average reduced power consumption, as well as 12% average perceived QoE gain.

Integrating the multitude of emerging internet of things (IoT) applications with diverse requirements in beyond fifth generation (B5G) networks necessitates the coexistence of enhanced mobile broadband (eMBB) and ultra-reliable low-latency communication (URLLC) services. However, bandwidth limited and congested sub-6GHz bands are incapable of fulfilling this coexistence. In this paper, we consider a reconfigurable intelligent surface (RIS)-aided wideband terahertz (THz) communication system to this end. In specific, we formulate a resource management problem, aiming at jointly optimizing the reflection coefficient of the RIS elements and the transmit power of the base station, as well as the wideband THz resource block allocation. To solve this problem, we adopt a supervised learning approach relying on optimization, deep learning and ensemble learning methods. Simulation results show that for an RIS of size$11\times 11$, up to 49% spectral efficiency gain is achieved for the eMBB service compared to the counterparts, while ensuring the reliability and latency requirements of the URLLC service. Further, the ensemble learning model can perform real-time resource management at the expense of up to 1% performance loss, compared to the optimization approach.

In this paper, the downlink transmission of a two-tier heterogeneous network (HetNet) is considered in which a macro base station (MBS) serves the macro users using orthogonal frequency division multiple access (OFDMA) and small base stations (SBSs) serve the small-cell users through multi-carrier non-orthogonal multiple access (MC-NOMA) and joint transmission (JT). In particular, assuming the subcarriers are already allocated to macro users, the problem of scheduling (i.e., joint user association and subcarrier allocation) and power control is studied with the goal of maximizing the total users’ perceived quality-of -experience (QoE) for small-cell users, while a minimum data rate for macro users is guaranteed. To solve the joint optimization problem, a near-optimal and computationally efficient two-phase solution approach is proposed based on the tools from optimization and machine learning (ML). In the first phase, the optimization problem is solved to obtain the scheduling decisions and transmit power variables. In the second phase, the optimized scheduling decisions and transmit power variables serve as training samples for an AlexNet classifier and support vector regressor (SVR), respectively. Simulation results reveal that the integration of JT into MC-NOMA, outperforms the conventional MC-NOMA scheme by up to 24%, 19%, and 21% for the web, video and audio multimedia services, respectively. Compared to a conventional convolutional neural network, our results demonstrate that for the web, video, and audio-services, AlexNet increases the scheduling prediction accuracy up to 14%, 11%, and 17%, while SVR increases the power prediction accuracy up to 8%, 7%, and 12%, respectively.

Passive beamforming in reconfigurable intelligent surfaces (RISs) enables a feasible and efficient way of communication when the RIS reflection coefficients are precisely adjusted. In this paper, we present a framework to track the RIS reflection coefficients with the aid of deep learning from a time-series prediction perspective in a terahertz (THz) communication system. The proposed framework achieves a two-step enhancement over the similar learning-driven counterparts. Specifically, in the first step, we train a liquid state machine (LSM) to track the historical RIS reflection coefficients at prior time steps (known as a time-series sequence) and predict their upcoming time steps. We also fine-tune the trained LSM through Xavier initialization technique to decrease the prediction variance, thus resulting in a higher prediction accuracy. In the second step, we use ensemble learning technique which leverages on the prediction power of multiple LSMs to minimize the prediction variance and improve the precision of the first step. It is numerically demonstrated that, in the first step, employing the Xavier initialization technique to fine-tune the LSM results in at most 26% lower LSM prediction variance and as much as 46% achievable spectral efficiency (SE) improvement over the existing counterparts, when an RIS of size 11×11 is deployed. In the second step, under the same computational complexity of training a single LSM, the ensemble learning with multiple LSMs degrades the prediction variance of a single LSM up to 66% and improves the system achievable SE at most 54%.

In this paper, we propose an intelligent analog beam selection strategy in a terahertz (THz) band beamspace multiple-input multiple-output (MIMO) system. First inspired by transfer learning, we fine-tune the pre-trained off-the-shelf GoogleNet classifier to learn analog beam selection as a multi-class mapping problem. Simulation results show 83% accuracy for the analog beam selection, which subsequently results in 12% spectral efficiency (SE) gain over the existing counterparts. For a more accurate classifier, we replace the conventional rectified linear unit (ReLU) activation function of the GoogleNet with the recently proposed Swish and retrain the fine-tuned GoogleNet to learn analog beam selection. It is numerically indicated that the fine-tuned Swish-driven GoogleNet achieves 86% accuracy, as well as 18% improvement in achievable SE, over the similar schemes. Eventually, a strong ensembled classifier is developed to learn analog beam selection by sequentially training multiple fine-tuned Swish-driven GoogleNet classifiers. According to the simulations, the strong ensembled model is 90% accurate and yields 27% gain in achievable SE in comparison with prior methods.

In this paper, we propose an accurate two-phase millimeter-Wave (mmWave) beamspace channel tracking mechanism. Particularly in the first phase, we train an extreme reservoir machine (ERM) for tracking the historical features of the mmWave beamspace channel and predicting them in upcoming time steps. Towards a more accurate prediction, we further fine-tune the ERM by means of Xavier initializer technique, whereby the input weights in ERM are initially derived from a zero mean and finite variance Gaussian distribution, leading to 49% degradation in prediction variance of the conventional ERM. The proposed method numerically improves the achievable spectral efficiency (SE) of the existing counterparts, by 13%, when signal-to-noise-ratio (SNR) is 15dB. We further investigate an ensemble learning technique in the second phase by sequentially incorporating multiple ERMs to form an ensembled model, namely adaptive boosting (AdaBoost), which further reduces the prediction variance in conventional ERM by 56%, and concludes in 21% enhancement of achievable SE upon the existing schemes at SNR = 15dB.

In this paper, we develop a resource allocation framework to optimize the downlink transmission of a backhaul-aware multi-cell cognitive radio network (CRN) which is enabled with multi-carrier non-orthogonal multiple access (MC-NOMA). The considered CRN is composed of a single macro base station (MBS) and multiple small BSs (SBSs) that are referred to as the primary and secondary tiers, respectively. For the primary tier, we consider orthogonal frequency division multiple access (OFDMA) scheme and also Quality of Service (QoS) to evaluate the user satisfaction. On the other hand in secondary tier, MCNOMA is employed and the user satisfaction for web, video and audio as popular multimedia services is evaluated by Quality-of-Experience (QoE). Furthermore, each user in secondary tier can be served simultaneously by multiple SBSs over a subcarrier via Joint Transmission (JT). In particular, we formulate a joint optimization problem of power control and scheduling (i.e., user association and subcarrier allocation) in secondary tier to maximize total achievable QoE for the secondary users. An efficient resource allocation mechanism has been developed to handle the non-linear form interference and to overcome the non-convexity of QoE serving functions. The scheduling and power control policy leverage on Augmented Lagrangian Method (ALM). Simulation results reveal that proposed solution approach can control the interference and JT-NOMA improves total perceived QoE compared to the existing schemes.