This paper introduces a deep reinforcement learning-based block coordinate descent (DRL-based BCD) algorithm to address the nonconvex weighted sum-rate maximization (WSRM) problem with a total power constraint. Firstly, we present an efficient block coordinate descent (BCD) method to solve the problem. While this method may not always achieve globally optimal solutions, it provides a pathway for integrating machine learning and domain-specific techniques with theoretical analysis of the underlying convexity of the subproblems. We then integrate deep reinforcement learning (DRL) techniques into the BCD method and propose the DRL-based BCD algorithm. This approach combines the data-driven learning capability of machine learning techniques with the navigational and decision-making characteristics of the optimization-theoretic-based BCD method. This combination significantly improves the algorithm’s performance by reducing its sensitivity to initial points and mitigating the risk of entrapment in local optima. The primary advantages of the proposed DRL-based BCD algorithm lie in its ability to adhere to the constraints of the WSRM problem and significantly enhance accuracy, potentially achieving the exact optimal solution. Moreover, unlike many pure machine-learning approaches, the DRL-based BCD algorithm capitalizes on the underlying theoretical analysis of the WSRM problem’s structure. This enables it to be easily trained and computationally efficient while maintaining a level of interpretability. Moreover, the DRL-based BCD framework demonstrates strong extensibility and can effectively be applied to other scenarios, such as joint beamforming for sum rate maximization, as demonstrated in this paper. Through numerical experiments, the DRL-based BCD algorithm demonstrates substantial advantages in effectiveness, efficiency, robustness, and interpretability for maximizing sum rates, which also provides valuable potential for designing resource-constrained AI-native wireless optimization strategies in next-generation wireless networks.

Continual semantic segmentation (CSS) has risen as a popular field, which aims to acquire new skills constantly without forgetting past knowledge catastrophically. In CSS, we identify that there is a severe imbalance between new classes and old classes, leading to the classifier weight toward new classes. In this paper, we deal with the continual semantic segmentation problem from the class imbalance perspective via mask-based class rebalancing, avoiding the model suffering from catastrophic forgetting. More specifically, the mask-based class rebalancing depends on a mask to combine resampling with reweighting ingenuously, which mitigates the classifier bias toward new classes. Besides, we also propose a frequency knowledge distillation, leveraging multiple frequency components information to maintain the feature representation space for old classes. We demonstrate the effectiveness of our approach with an extensive evaluation of the Pascal-VOC 2012 and ADE20K datasets, significantly outperforming the state-of-the-art method.

In this paper, we propose neural sum rate maximization, which is a neural network-based approach to tackle the nonconvex problem of maximizing the weighted sum rates with individual power constraints. Neural sum rate maximization combines both novel iterative optimization methods with data-driven models to deliver computationally efficient solution that learns the underlying statistics of the wireless network. Further-more, the solution can be refined by successive convex approximation and algorithm unrolling to accelerate the convergence of the neural sum rate maximization model training. We show that our algorithm is efficient for solving large-scale sum rate maximization problem. Numerical results validate the soundness and practicality of the proposed algorithm.

The COVID-19 pandemic brought not only global devastation but also an unprecedented infodemic of false or misleading information that spread rapidly through online social networks. Network analysis plays a crucial role in the science of fact-checking by modeling and learning the risk of infodemics through statistical processes and computation on mega-sized graphs. This article proposes MEGA, Machine Learning-Enhanced Graph Analytics, a framework that combines feature engineering and graph neural networks to enhance the efficiency of learning performance involving massive graphs. Infodemic risk analysis is a unique application of the MEGA framework, which involves detecting spambots by counting triangle motifs and identifying influential spreaders by computing the distance centrality. The MEGA framework is evaluated using the COVID-19 pandemic Twitter dataset, demonstrating superior computational efficiency and classification accuracy.

We propose an optimal scheduling strategy to enable fault-tolerant reliable computation to protect the integrity of computation. Specifically, we determine the optimal redundancy-failure rate tradeoff to incorporate redundancy into parallel computing units running multiple-precision arithmetic that are useful for applications such as asymmetric cryptography and fast integer multiplication. Inspired by network coding, we propose coding matrices to strategically map partial computation to available computing units, so that the central unit can reliably reconstruct the results of any failed machine without recalculations to yield the final correct computation output. We propose optimization-based algorithms to efficiently construct the optimal coding matrices subject to fault tolerance specifications. Performance evaluation demonstrates that the optimal scheduling effectively reduces the overall running time of parallel computing while resisting wide-ranging failure rates.