Covert communications have recently emerged as an innovative technology to enhance communication security and can utilize multiple antennas to improve system performance. Existing transmit precoding algorithms rely on the Gaussian inputs assumption, which is impractical in reality. Notably, conventional transmit precoding strategies for finite-alphabet inputs lack covertness considerations, limiting their effectiveness in ensuring covert communications. Therefore, based on finite-alphabet inputs, the practical transmit precoding for multiple-input multiple-output covert communication systems is investigated in this paper, where a warder equipped with multiple antennas aims to detect whether the transmission occurs. Specifically, an explicit expression for the outage probability of the legitimate communication is derived under imperfect channel state information. Then the covert throughput is formulated to measure transmission performance, while the Kullback-Leibler (KL) divergence is employed to evaluate covertness performance. Since both covert throughput and KL divergence are highly non-convex with respect to the transmit precoding matrix, maximizing covert throughput while satisfying the covertness constraint presents a significant challenge. To address the intractable optimization problem, an effective robust transmit precoding optimization algorithm based on deep reinforcement learning (DRL) is proposed to find the solution. Numerical results highlight the key differences between transmit precoding for Gaussian inputs and finite-alphabet inputs, while also demonstrating that the DRL algorithm outperforms other benchmark methods.

Public communication links can serve as shelters to facilitate covert transmissions. To improve the covertness performance, we regard multiple multi-antenna public users as friends and leverage their abundant spatial degrees of freedom (SDoF) to resist illegal surveillance. However, the resulting complex directional interference introduces challenges in measuring the covertness and optimizing the beamformers. Accordingly, this study investigates a covert communication system where multiple public user pairs coexist with one covert user pair in multiple-input multiple-output interference channels. Specifically, a covert rate maximization problem through joint transceiver beamforming is formulated, which is difficult to solve as it is a non-convex problem. To address this, a centralized algorithm is first developed based on the successive convex approximation method that transforms the problem into a convex optimization framework. Additionally, to alleviate the backhaul overhead, we develop a decentralized solution utilizing primal decomposition to decouple the interference management and the covertness metrics at each transmitter, relying only on local channel state information. Simulations demonstrate that our proposed algorithms fully exploit the SDoF jointly provided by the spatial distribution of public users and the employment of multi-antenna technology. Therefore, our algorithms achieve superior performance over baselines that do not effectively leverage these spatial resources.

The existing works on intelligent reflecting surface (IRS) aided covert communications consider the Gaussian input, which is however infeasible in practical systems. Two core issues remain to be answered: 1) How much performance gain can be obtained by applying the IRS for covert communications with finite-alphabet inputs? 2) How to jointly design the highly coupled parameters (constellation distribution and reflection coefficients) to obtain the optimal performance? To address these issues, in this work, the performance of the IRS aided covert communications with finite-alphabet inputs is analyzed, and a joint optimization scheme is proposed. In particular, the channel cutoff rate (CR) and the lower bound of the average detection error probability at the warder are derived under fading channels. Further, the impact of the reflection coefficients on the performance is discussed to reveal the benefits brought by the IRS. In addition, a two-layer algorithm is proposed to maximize the channel CR, where a channel variance tuple is introduced to decouple the optimization variables (constellation distribution and reflection coefficients). Numerical simulation demonstrates the superiority of the proposed scheme over various benchmarks. Moreover, the stricter the covertness constraint, the more concentrated the optimal probability distribution is at central constellation points.

The intelligent reflecting surface (IRS) has been widely applied in covert communications to hide the transmission behavior. However, the existing IRS relies on grid/battery power for its operation, which is unprocurable for practical covert communication applications. To address this issue, a novel multi-functional IRS (MF-IRS) is proposed for harvesting energy, signal reflecting, and amplifying simultaneously, where each element can flexibly switch between energy harvesting mode and passive/active reflection modes. To reveal the benefit of the MF-IRS for covert communications against multiple warders, the critical performance is analyzed, and effective design schemes are also provided. In particular, the detection error probabilities at warders are derived when the warders are non-collusive/collusive. In addition, the covert rate maximization problem is formulated by jointly optimizing the beamforming vector, element allocation matrices, and the reflection coefficient matrix. To solve this non-convex problem with highly coupled variables, an efficient successive convex approximation-based algorithm is proposed for the non-collusive scenario first and then extended to the collusive scenario. Simulation results demonstrate that the proposed MF-IRS always outperforms the self-sustainable passive/active IRSs, and it even outperforms the full-passive/active IRSs powered by grid/battery when the IRS is located near the signal source.

Accurately predicting protein-ligand affinity is one of the critical steps in the field of drug design. While deep learning approaches have shown great potential, existing methods based on sequence and 3D structural information still face challenges in capturing the spatial structure information and molecular bonding interactions between proteins and ligands. The advantage of molecular mechanics lies in its ability to account for complex molecular interactions, including electrostatic interactions and van der Waals forces. These interactions are key factors in determining the binding affinity between the pair of ligand and protein. Therefore, this study proposes a molecular mechanics-based heterogeneous graph attention neural network for predicting protein-ligand binding affinity. More specifically, various molecular mechanics features and atomic type features are first collected. Then, the heterogeneous graph is constructed for each pair of protein and ligand, in which the nodes are distinguished by atomic types, and the edges are categorized into covalent and non-covalent bonds. For graph representation learning, the molecular mechanics-guided aggregation mechanism is introduced to learn the meaningful constraints contained in the protein-ligand complexes. Finally, the representations of protein-ligand complexes are derived, on which the protein-ligand affinity is predicted accordingly. Experimental results show that the proposed model outperforms selected baselines on the task of protein-ligand affinity prediction.

Move structures have been studied in English for Specific Purposes (ESP) and English for Academic Purposes (EAP) for decades. However, there are few move annotation corpora for Research Article (RA) abstracts. In this paper, we introduce RAAMove, a comprehensive multi-domain corpus dedicated to the annotation of move structures in RA abstracts. The primary objective of RAAMove is to facilitate move analysis and automatic move identification. This paper provides a thorough discussion of the corpus construction process, including the scheme, data collection, annotation guidelines, and annotation procedures. The corpus is constructed through two stages: initially, expert annotators manually annotate high-quality data; subsequently, based on the human-annotated data, a BERT-based model is employed for automatic annotation with the help of experts’ modification. The result is a large-scale and high-quality corpus comprising 33,988 annotated instances. We also conduct preliminary move identification experiments using the BERT-based model to verify the effectiveness of the proposed corpus and model. The annotated corpus is available for academic research purposes and can serve as essential resources for move analysis, English language teaching and writing, as well as move/discourse-related tasks in Natural Language Processing (NLP).