In a communication system assisted by reconfigurable intelligent surfaces (RIS), phase optimization on RIS is not feasible by itself because of its passive nature. Therefore, the base station (BS) must convey the optimized phase shift to RIS via a dedicated control channel. However, the large number of RIS unit cells results in a considerable number of phase shift parameters, occupying a substantial amount of signaling resources. To address this challenge, this paper proposes a novel knowledge base autoencoder (KBAE) framework for compressing continuous phase shifts in RIS-aided communications. The KBAE framework integrates a specially designed phase shift compression network (PSCNet) and a higher compression ratio version (PSCNet-H) to achieve high-efficiency phase shift compression. The scheme utilizes a learnable knowledge base to approximate the distribution of RIS phase shift features. It only needs to transmit the indexes of the vectors in the knowledge base that are most similar to the RIS phase shift feature. Simulation results demonstrated that the proposed scheme can significantly improve the reconstruction accuracy of RIS phase shift and the average achievable rate compared to the benchmark methods. These advancements highlight the potential of KBAE to enhance the scalability and performance of future wireless networks.

Monotone chain polar codes generalize classical polar codes to multivariate settings and provide a flexible framework for achieving the entire admissible rate region of distributed lossless source coding. However, this flexibility introduces substantial challenges for successive cancellation (SC) decoding, as the decoding order may exhibit frequent variable switching that breaks the structural assumptions underlying classical implementations. In this paper, we develop a unified SC decoding framework for general monotone chain polar codes, accommodating arbitrary numbers of terminals, non-binary alphabets, and arbitrary monotone chains. We formulate the SC decoding as a series of inference subtasks over the polar transform and propose a computation graph framework based on probability propagation principles. Our analysis shows that the classical O(N) space optimization is not universally applicable to monotone chains, and that the decoding time complexity depends critically on the chain structure, ranging from O(NlogN) to O(N2). To address these challenges, we propose a novel iterative SC decoder that supports constant-time forking, together with a dedicated memory management strategy. The resulting framework enables time-efficient SC list decoding without relying on classical optimization techniques.

This paper investigates energy efficiency maximization for movable antenna (MA)-aided multi-user uplink communication systems by considering the time delay and energy consumption incurred by practical antenna movement. We first examine the special case with a single user and propose an optimization algorithm based on the one-dimensional (1D) exhaustive search to maximize the user's energy efficiency. Moreover, we derive an upper bound on the energy efficiency and analyze the conditions required to achieve this performance bound under different numbers of channel paths. Then, for the general multi-user scenario, we propose an iterative algorithm to fairly maximize the minimum energy efficiency among all users. Simulation results demonstrate the effectiveness of the proposed scheme in improving energy efficiency compared to existing MA schemes that do not account for movement-related costs, as well as the conventional fixed-position antenna (FPA) scheme. In addition, the results show the robustness of the proposed scheme to imperfect channel state information (CSI) and provide valuable insights for practical system deployment.

This paper explores the practical application of source polar codes to entropy coding tasks in modern transform coding pipelines. Transform coding remains the predominant and rapidly evolving framework for compressing complex real-world data. Despite the strong theoretical guarantees of polar codes, conventional polarization-based compression techniques follow a "construct-then-use" paradigm, which proves inefficient and inaccurate when applied to transform coding scenarios characterized by highly dynamic entropy models. To overcome this limitation, we propose a construction-free, plug-and-play polar compression scheme. Rather than relying on precomputed polarized entropies, our method selects output symbols based on probability vectors generated by a conditional entropy model. These vectors can be computed with low complexity and exact numerical precision, enabling efficient adaptation across diverse entropy coding tasks. The proposed approach offers greater flexibility than classical methods and achieves superior performance in the finite-length regime.

The growing demand for ultra-high-speed data transmission in short-reach optical interconnects exacerbates inter-symbol interference (ISI) and device-induced nonlinearities, presenting significant challenges for equalization. While deep neural networks offer strong modeling capabilities, their high complexity and limited adaptability hinder practical deployment in resource-constrained environments. To address these limitations, we propose the Dilated-Convolution Radial Basis Function network (DC-RBF), a lightweight neural equalizer designed to balance equalization performance and computational efficiency. Built on a fully convolutional backbone with a multi-symbol output structure, DC-RBF comprises three task-specific modules, each tailored to address distinct challenges: (i) non-causal dilated convolutions for efficient long-range ISI modeling, (ii) a radial basis function layer for fine-grained nonlinear compensation and robust cross-link adaptability, and (iii) enhanced residual connections that strengthen feature fusion and improve training stability. Experiments on an 850 nm VCSEL-MMF IM/DD platform demonstrate that DC-RBF enables PAM-8 transmission at 165 Gbps (back-to-back) and 159 Gbps (100 m MMF) within the 20% soft-decision FEC threshold, achieving strong generalization without model retraining. With low computational overhead (fewer than 1000 multiply-accumulate operations), DC-RBF achieves superior BER performance, while reducing inference latency by 28% and model size by up to 97% compared to deeper neural baselines. These results validate the effectiveness and deployment potential of DC-RBF for next-generation optical interconnects.

In future wireless communication systems, reconfigurable intelligent surfaces (RIS) present a promising technology. However, in RIS-assisted communication systems, feedback of downlink channel state information (CSI) poses significant challenges, particularly with a large number of base station (BS) antennas and RIS unit cells. Traditional CSI feedback methods based on compressed sensing assume channel sparsity and require extensive computation and storage operations by both the user equipment (UE) and the BS, thereby increasing the burden on the UE. This paper proposes a lightweight CSI feedback mechanism based on global context attention network (GCANet). This method uses a lightweight autoencoder at the UE for CSI encoding, reducing the computational and storage burden on the UE, while deploying the decoder at the BS to leverage its powerful computational capabilities for complex decoding tasks. Simulation results demonstrate that this method significantly reduces feedback overhead and enhances system performance, with the UE's encoder model parameters and computational load being substantially reduced compared to baseline methods.

Classical source polar codes require the construction of frozen sets for given sources. While this scheme offers excellent theoretical performance, it faces challenges in practical data compression systems, including sensitivity to the accuracy and computational complexity of the construction algorithm. In this letter, we explore the feasibility of construction-free polar compression schemes. By optimally selecting output symbols based on the decoder's behavior, the proposed scheme not only enhances flexibility but also achieves significant improvements in compression rates. Several enhancements are introduced to facilitate the practical implementation of the proposed scheme. Numerical results demonstrate the superior performance compared to existing polar compression approaches.

This paper proposes an optics-informed residual convolution network with learnable activation function for high-speed VCSEL-MMF optical interconnects. The proposed method significantly enhances performance across 130–140 Gbps PAM-4 transmissions, outperforming Volterra and conventional neural network baselines.

This paper investigates the energy efficiency optimization for movable antenna (MA) systems by considering the time delay and energy consumption introduced by MA movement. We first derive the upper bound on energy efficiency for a single-user downlink communication system, where the user is equipped with a single MA. Then, the energy efficiency maximization problem is formulated to optimize the MA position, and an efficient algorithm based on successive convex approximation is proposed to solve this non-convex optimization problem. Simulation results show that, despite the overhead caused by MA movement, the MA system can still improve the energy efficiency compared to the conventional fixed-position antenna (FPA) system.

Fiber nonlinearity compensation is crucial for enhancing the transmission distance and data capacity of wavelength-division multiplexing (WDM) coherent optical communication systems, while existing equalizers based on conventional neural networks face a fundamental limitation of prohibitive computational complexity. In this Letter, we propose a gradient-driven pruned Kolmogorov-Arnold network (GDP-KAN) for efficient fiber nonlinearity compensation. Leveraging learnable spline activation functions, the KAN architecture achieves superior performance with a lightweight structure. Furthermore, we introduce a gradient-driven pruning strategy based on attribution score, which sparsifies the network according to the signal characteristics of nonlinear impairments, enabling ultralow-complexity operation. We carry out an experiment of 8-channel WDM transmission over 1600 km standard single-mode fiber (SSMF) with 64 GBaud polarization-division-multiplexed (PDM) 16-ary quadrature amplitude modulation (16-QAM) signals. The proposed GDP-KAN requires only 300 real multiplications per bit (RMPB), outperforming 1 step per span DBP by 0.63 dB Q 2 factor gain with merely 12% of its complexity, while reducing the RMPB by 68.75% compared to the multilayer perceptron (MLP) based equalizer without performance degradation.

Against the backdrop of the vigorous development of artificial intelligence (AI) large models, the demand for the data center optical module rate continues to rise. To meet the requirements of high rate, low power consumption, low latency, and high integration for ultra-high-speed short-reach optical interconnect, the development trend of the optical module industry was introduced, and relevant equalization and coding technologies were discussed. In terms of equalization technology, a low-complexity Volterra equalization method based on the least angle regression strategy (LaNLE) and a deep neural network equalization method based on hidden feature extraction (HFE) were proposed. Under the 120 Gbit/s PAM-8 signal transmission, the complexity of LaNLE was reduced by 70.1% compared with traditional methods at the same bit error rate; HFE improved the neural network training efficiency and achieved 288 Gbit/s PAM-8 signal transmission. In terms of coding technology, the optimization method of QC-LDPC codes in the optical interconnection system and the LDPC joint equalization and decoding method based on deep learning were introduced. Experimental results show that the proposed methods can effectively optimize the bit error rate and improve the system performance. At the same time, the future research directions were prospected.

This letter begins with a comparative analysis of the error performance between co-frequency co-time full-duplex (CCFD) and half-duplex (HD) systems in the finite blocklength regime for a given bandwidth. It is demonstrated that the CCFD system provides significant advantages, not only by increasing data rates through the doubled bandwidth for bidirectional transmissions, but also by allowing a larger number of transmitted data symbols, thereby enabling the longer blocklengths, which can reduce the error rate further. Subsequent analyses focus on the issue of spectral efficiency for a given error probability. Notably, gains exceeding 2 are observed when the self-interference (SI) is effectively canceled below a certain threshold. Finally, numerical evaluations substantiate the theoretical findings presented in this study.

This paper proposes an optics-informed fully convo-lutional network equalizer for high-speed VCSEL-based optical interconnects. The proposed method demonstrates significant performance improvements across different transmission rates and received optical powers.

Full-duplex technology can improve bandwidth and energy efficiency but has serious self-interference. In this paper, we introduce a novel full-duplex communication scheme that leverages the Doppler effect. By rotating the circular antenna array, the Doppler frequency shift is introduced into the received uplink (UL) signal, and the UL signal can be separated from the downlink (DL) signal in the frequency domain, thus eliminating self-interference. In order to maximize the interference-free bandwidth, an antenna switching criterion to maximum frequency offset interval is proposed for antenna switching control. Moreover, an antenna switching module based on the above criterion and a Doppler frequency shift compensation module are designed. Simulation results show that the proposed scheme can effectively avoid interference between UL and DL signals and achieves a substantial capacity improvement compared to conventional systems.

Semantic Communication (SemCom) is envisaged as the next-generation paradigm to address challenges stemming from the conflicts between the increasing volume of transmission data and the scarcity of spectrum resources. However, existing SemCom systems face drawbacks, such as low explainability, modality rigidity, and inadequate reconstruction functionality. Recognizing the transformative capabilities of AI-generated content (AIGC) technologies in content generation, this paper explores a pioneering approach by integrating them into SemCom to address the aforementioned challenges. We employ a three-layer model to illustrate the proposed AIGC-assisted SemCom (AIGC-SCM) architecture, emphasizing its clear deviation from existing SemCom. Grounded in this model, we investigate various AIGC technologies with the potential to augment SemCom's performance. In alignment with SemCom's goal of conveying semantic meanings, we also introduce the new evaluation methods for our AIGC-SCM system. Subsequently, we explore communication scenarios where our proposed AIGC-SCM can realize its potential. For practical implementation, we construct a detailed integration workflow and conduct a case study in a virtual reality image transmission scenario. The results demonstrate our ability to maintain a high degree of alignment between the reconstructed content and the original source information, while substantially minimizing the data volume required for transmission. These findings pave the way for further enhancements in communication efficiency and the improvement of Quality of Service. At last, we present future directions for AIGC-SCM studies.

This letter investigates the challenges encountered by deep joint source-channel coding in erasure channels. We explore the effectiveness of the widely adopted dropout technique in endowing deep neural networks with resilience against erasures. However, directly applying dropout at the channel layer introduces uncertainty into the neural network’s training process, leading to performance degradation. To address this issue, we introduce the Polarized Dropout scheme and a novel network architecture that encodes analog symbols using the Walsh-Hadamard transform based on real-number field computation. Leveraging the polarization of symbol recovery probabilities, for a given erasure rate, a determined set of neurons will be assigned a dropout rate of 1, while the remainder are assigned a dropout rate of 0. Simulation results indicate a maximum enhancement of nearly 6dB in communication performance.

The rapidly evolving field of generative artificial intelligence technology has introduced innovative approaches for developing semantic communication (SemCom) frameworks, leading to the emergence of a new paradigm—generative AI-assisted SemCom (GSC). Benefiting from its strong ability to understand and generate high-quality content across various domains, this approach can effectively address the reconstruction limitations that challenge traditional SemCom systems. However, this architecture often suffers from high latency due to the complex processes involved in semantic extraction and generative semantic inference. To mitigate this issue, we propose a low-latency GSC framework, achieved by enabling the parallel execution of the transmitter’s semantic extracting and the receiver’s generating processes from a macro perspective. Furthermore, to attain more accurate and semantically aligned reconstruction, we design a temporal prompt engineering approach that utilizes reinforcement learning to sequence the temporal feature extraction steps at the transmitter. The results show that compared to the conventional GSC architecture, our designed framework can achieve a 52% reduction in residual task latency that extends beyond the fixed inference duration while only incurring an approximate 9% decrease in task score.

In this paper, we innovatively associate the mutual information with the frame error rate (FER) performance and propose novel quantized decoders for polar codes. Based on the optimal quantizer of binary-input discrete memoryless channels (B-DMCs), the proposed decoders quantize the virtual subchannels of polar codes to maximize mutual information (MMI) between source bits and quantized symbols. The nested structure of polar codes ensures that the MMI quantization can be implemented stage by stage. Simulation results show that the proposed MMI decoders with 4 quantization bits outperform the existing nonuniform quantized decoders that minimize mean-squared error (MMSE) with 4 quantization bits, and yield even better performance than uniform MMI quantized decoders with 5 quantization bits. Furthermore, the proposed 5-bit quantized MMI decoders approach the floating-point decoders with negligible performance loss.

Semantic communication (SemCom) has emerged as a promising architecture in the realm of intelligent communication paradigms. SemCom involves extracting and compressing the core information at the transmitter while enabling the receiver to interpret it based on established knowledge bases (KBs). This approach enhances communication efficiency greatly. However, the open nature of wireless transmission and the presence of homogeneous KBs among subscribers of identical data type pose a risk of privacy leakage in SemCom. To address this challenge, we propose to leverage the simultaneous transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) to achieve privacy protection in a SemCom system. In this system, the STAR-RIS is utilized to enhance the signal transmission of the SemCom between a base station and a destination user, as well as to convert the signal to interference specifically for the eavesdropper (Eve). Simulation results demonstrate that our generated task-level disturbance outperforms other benchmarks in protecting SemCom privacy, as evidenced by the significantly lower task success rate achieved by Eve.

We study scheduling algorithms to minimize the Age of Information (AoI) regrets under non-stationary channels in a drifting environment, where the probability that a channel can successfully transmit a packet varies over time and the total variation of probabilities across adjacent time slots is bounded. We characterize the AoI regret lower bound for single-source and multi-source systems, and prove that a restarted channel scheduling algorithm achieves an AoI regret within a logarithmic factor gap from the lower bound for a single-source system. We augment the above channel scheduling algorithm with a source scheduling algorithm to support multiple source transmissions where sources are decoupled from channels. In addition, we develop a scheduling algorithm for multiple source transmissions where sources and channels are coupled, and characterize the AoI regret upper bounds.