Compared to Arikan’s G2 kernel, large-kernel polar codes exhibit higher polarization rates and superior error correction performance. The critical steps of exact successive cancellation (SC) decoding for such codes can be implemented via trellis-based computations to reduce complexity. However, the complexity remains high for large kernels. This paper proposes a permutation-based trellis optimization scheme. The approach builds on the Massey minimal trellis and reorders its time axis to find a permutation that minimizes the number of trellis edges, thereby further reducing the exact SC decoding complexity. For smaller kernels (G3–G12), an exhaustive search is conducted to identify the optimal trellis. For larger kernels (G13–G16), where an exhaustive search becomes infeasible due to the factorial growth of the permutation space, an ant colony optimization (ACO)-based method is employed to find a near-optimal permutation. Simulation results show that the permutation-optimized trellis lowers the direct SC decoding complexity drastically. Furthermore, compared to the l-expression, the W-formula and original Massey trellis methods, it achieves multiplication operation reductions of up to 99.2%, 58.1%, and 56.5%, respectively. The improvement is particularly beneficial for large kernels, where traditional decoding methods become computationally prohibitive.

Multi-kernel polar codes, which are constructed by different sizes of kernels make the length of the traditional polar code more flexible. In this paper, we extend the multi-kernel polar codes to 5x5 kernels including kernel design, code design, and fast decoding. First, we construct a new 5 x 5 kernel with the maximum polarization rate. The first row of this kernel is an all-one vector, which enable us to find the relation between node modification and distance property of multi-kernel polar codes. Second, we modify the repetition (REP) nodes to the new 5 x 5 kernel, which can increase the distance property of the constructed polar codes, improve the decoding performance and reduce memory requirements during the fast decoding. In addition, we find our REP node modification can be generalized to multi-kernel polar codes when the first row of the used kernels is all-one vector. Third, we prove the four kinds of basic nodes (Rate-1 node, Rate-0 nodes, Repetition (REP) nodes, and Single-Parity check (SPC) nodes) of fast decoding are suitable for the new kernel. Moreover, we generalize the Generalized Parity-Check (G-PC) nodes and the Generalized Repetition (G-REP) nodes, which are suitable for multi-kernel polar codes constructed by any size of kernels and any order of kernels. By simulations, we show that our modification improves the error-rate performance of the fast decoding. Meanwhile, the fast decoding of our multi-kernel polar codes has at least 68.4% latency reduction compared with the traditional SC decoder in all cases considered where code lengths are 80, 135, 400, and 675.

Current Retrieval-augmented-generation (RAG) in Open-domain question-answering (ODQA) relies heavily on the standardization of question expressions and the quality of retrieved passages and this heavy reliance leads to a decline in the accuracy of ODQA results. To address this problem, the question and passage augmentation (QP-aug) method proposes a question augmentation module and a passage self-generation module based on LLMs, which enhance the accuracy of ODQA results. However, the QP-aug method has certain limitations in multi-hop questionanswering (QA) scenarios because it is a single-step RAG strategy. To enhance the universality of the QP-aug method for ODQA, we propose an iterative QP-aug (IQP-aug) framework and introduce an intelligent iteration control mechanism, including the verification-evaluation dual gating (VEDG) module and the bestprediction selector (BPS) module. The VEDG and BPS modules enable adaptive selection of multi-granular iteration steps, thereby generating the optimal iteration strategy for single-hop and multihop questions. Our approach enhances the framework’s universality, improves result accuracy, and saves computing resources. Experimental results show that IQP-aug improves F1 by 12.2% on the single-hop dataset and 30.8% in the multi-hop dataset compared with the baseline method, verifying the effectiveness of the IQP-aug framework in adapting to single-hop and multi-hop question types.

In this letter, we optimize the weight distribution of the polar codes by exchanging elements in the information set and the frozen set. Based on the work of M. Rowshan et al., we propose two improvements. First, the rows with the twice weight of the minimum weight row are usually selected by the original scheme for the added rows. By our improvement, we select rows with the largest weights such as 4 or 8 times of the minimum weight row. This modification optimizes the weight distribution of polar codes. Furthermore, if there is a case where the selected row weights are the same, the row with higher reliability is selected. Second, we consider the influence of balancing rows for the removed rows. In the original scheme, one core row is removed and then an added row is selected at each step. Our strategy is to fix the position of several added rows, and then remove the core rows or balancing rows globally. This modification can not only reduce the number of Min-Weight codewords but also optimize the channel reliability of the removed rows. Simulation results show that the decoding performance of modified codes has been greatly improved at the practical block error rate of 10-2-10-3 without changing the code rate of the polar codes.

Retrieval-Augmented Generation (RAG) is a key technique for enhancing Large Language Models (LLMs), yet state-of-the-art systems often rely on computationally expensive, multi-pass inference strategies like Self-Consistency to achieve high accuracy, limiting their practical applicability. This paper challenges a critical, yet overlooked, design flaw in existing efficient RAG frameworks: the suboptimal ordering of retrieved context that fails to align with known positional biases of LLMs. To address this, we propose Efficient-RAG (ERAG), a novel framework that operates in a single pass by strategically reordering context via a "Relevance-First" strategy. Experiments on Marathon and NovelQA show that our method dramatically improves the performance of smaller models (1B, 3B). Crucially, our single-pass approach even outperforms a costly 10-pass baseline in key scenarios, while improving inference efficiency by approximately 10x across all settings. On larger models (8B), ERAG achieves highly competitive accuracy, representing a superior performance-cost trade-off. Our work not only presents an efficient alternative to costly voting mechanisms but also provides a key insight: the optimal context strategy is model-scale dependent, paving the way for more adaptive and resource-aware RAG systems.

Multi-kernel polar codes, which are constructed by different sizes of kernels make the traditional polar codes have more flexible code length. In this paper, we extend the multi-kernel polar codes to 5x5 kernels including kernel design, code design, and fast decoding. First, we construct a new 5x5 kernel, which has the maximum polarization rate. The first row of this kernel is all-one vector, which enable us to seek the relation between node modification and distance property of multi-kernel polar codes. Second, we modify the repetition (REP) nodes to this kernel, which can increase the distance property of the constructed polar codes, improve the decoding performance and reduce memory requirements during the fast decoding. In addition, we find this REP node modification can be generalized when multi-kernel polar codes constructed by kernels, which the first row is all-one vector. Third, we prove the four kinds of basic nodes (Rate-1 node, Rate-0 nodes, Repetition (REP) nodes, and Single-Parity check (SPC) nodes) of fast decoding are suitable for this kernel. Moreover, we generalize the Generalized Parity-Check (G-PC) nodes and the Generalized Repetition (G-REP) nodes, which are suitable for multi-kernel polar codes constructed by any size of kernels and any order of kernels. The G-PC and G-REP nodes can further accelerate the decoding speed. By simulations, we show that our modification improves the error-rate performance of the fast decoding. Meanwhile, the fast decoding of our multi-kernel polar codes has at least 68.4% reduced latency compared with the traditional SC decoder in all cases considered where frame lengths are (80, 135, 400, 675).

The Polar-Adjusted Convolutional (PAC) codes, improve the error correction ability of the polar code by concatenating convolutional transformation and polar transformation. For the traditional stack decoding algorithm, the ideal performance has not yet been achieved for some PAC codes. In this paper, we adopt a multi-stack decoding algorithm to further improve the error performance of PAC codes. In the multi-stack decoding algorithm, the original single stack is converted into multiple stacks, which expands the exploration path of decoding, obtains more possible decoding results, and improves the decoding performance of PAC codes. Compared to the single-stack decoding, the multi-stack decoding achieves a performance gain of approximately 0.2dB for a PAC code with code length 128 and rate 0.5.

In this paper, the reciprocal channel approximation (RCA) method is used to design polar codes with large kernels. Firstly, based on the given large kernels, we obtain the encoder graph. Secondly, the bit-channel SNR parameters are successively updated according to the encoder graph by the RCA. Finally, the frozen bits are selected by the SNR of the last bit-channels. The RCA method can avoid the distortion caused by the kernel matrix being too large. Furthermore, in the calculation process, the closed-form expression of the channel capacity is used to avoid excessive use of transcendental functions, which can greatly reduce the calculation cost. The experimental results show that for the G3⊗5 and G5⊗3 with lengths of 243 and 125, and bit rates of 1/2, when the channel SNR is more than 3dB and 2.5dB, respectively, polar codes with large kernels designed by our scheme have better frame error rate than the Gaussian approximation (GA) method.

Polar codes constructed by large kernels can attain better finite length performance than those originating from Arıkan’s $2\times 2$ kernel. However, the successive cancellation (SC) decoding for these polar codes is impractical even for relatively small kernel size of $m$ because complexity of the kernel computation grows exponentially with $m$ . This research shows when $m>2$ , there exists a large amount of like terms in the kernel computation which yields a ground for facilitating the decoding. By transferring the kernel computation from the probability domain to the likelihood ratio domain ( $l$ -domain), the so- called $l$ -formula method provides an efficient way to combine the like terms in the kernel computation for kernels up to size 11. However, the $l$ -formula method becomes intractable for kernel size beyond 11. To further reduce the computational complexity, this paper proposes a $W$ -formula method which transforms the kernel computation into the probability pair domain ( $W$ -domain). Advanced from the $l$ -domain, the numerator and denominator of the likelihood ratio are considered separately, which eases the restrictions of combining like terms. The $W$ -formula method can combine much more like terms resulting in a significant reduction on the number of sub-formulas for medium kernels ( $m\leq 16$ ). Furthermore, in the $W$ -domain, sub-formulas become regular and there exist many common sub-formulas whose computations can be shared. Being able to handle kernels of size up to 16, we show that the $W$ -formula based SC decoding achieves a significant complexity reduction over the existing non-approximate SC decoding (the $l$ -formula based SC decoding).

A trellis decoding scheme based on Massey trellis for polar codes with a ternary kernel is proposed. First, we construct trellises for kernel computations of the successive cancellation (SC) decoder based on the Massey trellis. Then, the scheme calculates the bit-channel transition probability through these trellises, which provides an appropriate framework to take advantage of the distributive law and can effectively reduce the number of operations. Simulation results show that for a G5⊗3 polar code of length 243 and rate 0.5, our scheme can save 14.2% of the computational cost with no sacrifice in error performance under the SC decoder.

With the development of intelligent monitoring and autonomous driving technology, the adverse effects of rainy weather are receiving increasing attention. The current image de-raining techniques have two main drawbacks: first, they easily lose local features and channel features, and second, the computational cost of global calculations on large feature maps is staggering. In this paper, we propose a self-attention network called URformer, which integrates spatial features and channel features. We introduce a novel feature extraction module that effectively combines local self-attention and channel self-attention mechanisms. URformer significantly reduces the computational cost of traditional global Transformers while extracting and preserving channel and spatial features. Experimental results demonstrate that the URformer network exhibits excellent performance on the RGB image de-raining task.

As the largest target in remote sensing images, buildings have important application value in urban planning and old city reconstruction. However, most networks have poor recognition ability on high resolution images, resulting in blurred boundaries in the segmented building maps. Then, the similarity between buildings and backgrounds will lead to inter-class indistinction. Finally, the diversity of buildings brings difficulties to segmentation, which requires the network to have better generalization ability. To address these problems, we propose Fusion Multi-scale Attention Mechanism Network (FMAM-Net). Firstly, we design Feature Refine Compensation Module(FRCM) to improve the boundary ambiguity problem, including Feature Refinement Module(FRM) and Feature Compensation Module(FCM). FRM utilizes the densely connected architecture to refine features and increase recognition capabilities. FCM introduces low-level features to make up for the lack of boundary information in high-level features. Secondly, to handle inter-class indistinction, we design Tandem Attention Module(TAM) and Parallel Attention Module(PAM). TAM is designed to sequentially filter some features from channels and spaces for adaptive feature refinement. PAM combines context information and uses high-level features to guide low-level features to select more distinguishable features. Finally, based on the binary cross entropy loss function, we add an evaluation index to reduce the error caused by determining the optimization direction only through cross entropy. On the Inria Aerial Image Labeling Dataset, FMAM-Net achieves mean IoU of 85.34%, which is 5.58% higher than AMUNet and 3.77% higher than our baseline(U-Net ResNet-34). On the WHU Dataset, IoU reached the maximum value of 91.06% on FMAM-Net, 1.67% higher than SARB-UNet and 0.2% higher than MAP-Net. The visualization results show that FMAM-Net improves the fuzzy boundary of building segmentation and reduces the inter-class indistinction.

专业培养方案是学生培养的规范和标准,课程体系是专业培养方案的主体之一,构建基于OBE理念,足以支撑毕业要求和培养目标的专业课程体系是为社会输送合格的工程教育人才,实现新工科建设,建成工程教育强国,形成中国特色世界一流工程教育体系的前提.文章基于浙江师范大学电子信息工程专业工程教育专业认证经验和实践,介绍了电子信息工程专业课程体系,各课程与认证标准的关系,各模块对学生能力的培养及与解决复杂工程问题能力培养的支撑关系.为同类专业修订符合工程教育专业认证要求的课程体系提供参考.

In order to construct high-dimensional core matrix polar codes quickly,a two-stage polar code construction method TPMC based on Monte Carlo(MC) is proposed.In the first stage,the most reliable and unreliable bits are obtained by using the Gauss approximation method with linear complexity.In the second stage,the above bits are fixed as frozen bits and MC method is implemented to measure the error rate of the remaining bit channels.The bits with a lower error rate are selected from the remaining bits,and are combined with most reliable information bits in the first stage to form an information bits set.Simulation results show that compared with MC method,TPMC method can reduce the computational complexity and improve the decoding efficiency.

提出一种基于修改的连续消去(MSC)译码算法的极化码时延降低方法.在MSC译码算法中,可以降低类型-I和类型-II节点的时延.通过重新分配信息位,所提出的方法可以获得一种较好的类型-I和类型-II节点的分布,在MSC译码算法下可以进一步降低译码时延.仿真结果表明,在与原算法的译码性能相比可忽略的损失下,由该方法构造的极化码可以达到8.5％的时延降低.而且,该方法易于调整差错性能与译码时延之间的权衡.

We propose a two-phase Monte Carlo (TPMC) method to accelerate the original Monte Carlo (MC) method for constructing polar codes with high-dimensional kernels. In the TPMC method, some of the most reliable and unreliable bits are obtained by Gaussian approximate-density evolution (GA-DE) method in the first phase; in the second phase, these most reliable and unreliable bits are viewed as frozen bits. Then, the MC method are used to evaluate the remaining bits and select some best bits from the remaining bits. Finally, these best bits and the most reliable bits are combined as the information bits of the constructed polar code. By our investigation, most bits can be fixed as frozen bits in the second phase without error performance loss of the constructed polar codes. Because computation of frozen bits can be saved and computation of the GA-DE method can be ignored in contrast to that of the MC method, the TPMC method substantially reduces the complexity of the MC method. Simulation results show that 1) For a $G_{15}^{\otimes 3}$ polar code with block length 3375 and code rate 1/2, the TPMC method can fix 3200 bits as frozen bits in the second phase without error performance loss, which reduces the computation cost by approximately 92.6% over the MC method; 2) Polar codes with high-dimensional kernels constructed by the TPMC method outperforms polar codes with the $G_{2}$ kernel constructed by the Tal-Vardy method in terms of error performance; 3) With the same computational cost, the TPMC method can construct better polar codes than the MC method.

The authors have retracted this conference chapter [1] because it shows significant overlap with a previously published chapter [2]. All authors agree to the retraction. [1] Chen C., Dai J., Cheng C., Huang Z. (2018) A Resource Allocation Algorithm Based on Game Theory in UDN. In: Gu X., Liu G., Li B. (eds) Machine Learning and Intelligent Communications. MLICOM 2017. Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, vol 226. Springer, Cham [2] Y. Liu, Y. Wang, Y. Zhang, R. Sun and L. Jiang, “Game-theoretic hierarchical resource allocation in ultra-dense networks,” 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), Valencia, 2016, pp. 1–6. https://doi.org/10.1109/PIMRC.2016.7794819

This paper presents a new latency reduction method for successive-cancellation (SC) decoding of polar codes that performs a frozen-bit checking on the rate-other (R-other) nodes of the Fast Simplified SC (Fast-SSC) pruning tree. The proposed method integrates the Fast-SSC algorithm and the Improved SSC method (frozen-bit checking of the R-other nodes). We apply a recognition-based method to search for as many constituent codes as possible in the decoding tree offline. During decoding, the current node can be decoded directly, if it is a special constituent code; otherwise, the frozen-bit check is executed. If the frozen-bit check condition is satisfied, the operation of the R-other node is the same as that of the rate-one node. In this paper, we prove that the frame error rate (FER) performance of the proposed algorithm is consistent with that of the original SC algorithm. Simulation results show that the proportion of R-other nodes that satisfy the frozen-bit check condition increases with the signal-to-noise-ratio (SNR). Importantly, our proposed method yields a significant reduction in latency compared to those given by existing latency reduction methods. The proposed method solves the problem of high latency for the Improved-SSC method at a high code rate and low SNR, simultaneously.

Large-scale multiple-input multiple-output (MIMO) system has the advantages of high energy efficiency and spectrum utilization. But using some cheap hardware may cause some problems, such as nonlinearity of the high power amplifier (HPA). When HPA works in the nonlinear region, it will affect the received signal and greatly reduce the performance of the system. In this paper, we first study the impact caused by nonlinear HPA, and then we optimize the traditional precoding algorithm to design an improved precoding algorithm which can reduce the impact. The simulation results show that the proposed algorithms perform better in bit error ratio and system capacity compared to the block of diagonalization (BD) precoding algorithm and forced zero (ZF) precoding algorithm, especially in the condition of high signal to noise ratio (SNR). So we can draw the conclusion that the algorithms proposed in this paper are able to reduce the impact caused by nonlinear HPA to the system.