Qiuyan Wang

Graph Convolutional Networks (GCNs) enhance node representations by aggregating information from neighboring nodes, but deeper layers often suffer from over-smoothing, where node embeddings become indistinguishable. Transformers enable global dependency modeling on graphs but suffer from high computational costs and can exacerbate over-smoothing when multiple attention layers are stacked. To address these issues, we propose GLADC, a novel framework designed for semi-supervised node classification. It integrates global linear attention for efficient long-range dependency capture and a dual constraint module for local propagation. The dual constraint consists of (1) column-wise random masking on the representation matrix to dynamically limit redundant information aggregation, and (2) row-wise contrastive constraint to explicitly increase inter-node distance and preserve distinctiveness. This design achieves linear-complexity global mixing while effectively countering representation homogenization. Extensive evaluations on seven real-world datasets demonstrate that GLADC delivers competitive performance and maintains robustness in deep architectures (up to 32 layers). An ablation study further confirms the synergistic effect of both constraints in alleviating over-smoothing and preventing premature convergence.

Codebooks possessing optimal correlation property are utilized to differentiate signals among various users in code-division multiple-access (CDMA) systems. In this paper, a class of codebooks is proposed based on additive characters over finite fields and is verified to be optimal with respect to the Levenshtein bound. By shortening the length of the presented optimal codebooks, a kind of asymptotically optimal codebooks is obtained. To our knowledge, the presented asymptotically optimal codebooks have new parameters. Further more, these optimal and asymptotically optimal codebooks are highly desirable in synchronous Direct-Sequence CDMA systems, as they facilitate the expansion of subscriber capacity within the constrained spectral resources.

The advancement of deep learning has driven extensive research validating the effectiveness of U-Net-style symmetric encoder-decoder architectures based on

With the development of artificial intelligence technology, the demand for multi-domain pose estimation in complex scenes is growing. The existing single model solution faces the problem of insufficient ability in cross domains of multi-scene, which is caused by the different numbers of keypoints and complex features of samples. To solve this problem, we propose a new method that has a four-stage pose estimation framework. This framework applies methods such as object detection, domain classification and pose estimation. As the experimental results show, on the testing set of mixed domains, the accuracy of our method is 5.1% higher than the best one of the existing methods, which ensures high performance pose estimation in many applications.

Cell length is one of important parameters of cell morphology, which is of great significance for understanding cell structure, function,and physiological status. This paper uses the Omnipose method to segment rod-shaped bacterial cells, determine their skeleton and measure the length of rod-shaped bacterial cells. Compared with other algorithms, this algorithm proposed in this paper has smaller errors, achieving higher accuracy and stability in measuring bacterial cell length. The method described in the paper can quickly and automatically batch process a large number of cell images, and calculate the length of cells with high accuracy. This algorithm also provides a convenient, efficient, and reliable method for the study and experimentations of cell morphology.

In this paper, we give a construction of strongly regular Cayley graphs on the finite field F _qⁿ . As applications of these strongly regular Cayley graphs, a class of codebooks is presented and proved to be asymptotically optimal with respect to the Welch bound. Further more, these constructed codebooks have new parameters.

LSB (Least Significant Bit) sequences are widely used as the initial inputs in some modern stream ciphers, such as the ZUC algorithm-the core of the 3GPP LTE International Encryption Standard. Therefore, analyzing the statistical properties (for example, autocorrelation, linear complexity and 2-adic complexity) of these sequences becomes an important research topic. In this paper, we first reduce the autocorrelation distribution of the LSB sequence of a p-ary m-sequence with period pn – 1 for any order n≥2 to the autocorrelation distribution of a corresponding Costas sequence with period p – 1, and from the computing of which by computer, we obtain the explicit autocorrelation distribution of the LSB sequence for each prime p<100. In addition, we give a lower bound on the 2-adic complexity of each of these LSB sequences for all primes p<20, which proves to be large enough to resist the analysis of RAA (Rational Approximation Algorithm) for FCSRs (Feedback with Carry Shift Registers). In particular, for a Mersenne prime p=2k – 1 (i.e., k is a prime such that p is also a prime), our results hold for all its bit-component sequences since they are shift equivalent to the LSB sequence.

In this paper, a family of [q2-1,4,q2 -q-2] cyclic codes over 픽q meeting the Griesmer bound is presented. Their duals are [q2-1,q2-5,4] almost MDS codes and are optimal with respect to the sphere-packing bound. The q0-ary subfield codes of this family of cyclic codes are also investigated, where q0 is any prime power such that q is power of q0. Some of the subfield codes are optimal and some have the best known parameters. It is shown that the subfield codes are equivalent to a family of primitive BCH codes and thus the parameters of the BCH codes are solved. The duals of the subfield codes are also optimal with respect to the sphere-packing bound. A family of [q2,4,q2-q-1] linear codes over 픽q meeting the Griesmer bound is presented. Their duals are [q2,q2-4,4] almost MDS codes and are optimal with respect to the sphere-packing bound. The q0-ary subfield codes of this family of linear codes are also investigated, where q0 is any prime power such that q is power of q0. Five infinite families of 2-designs are also constructed with three families of linear codes of this paper.