Electroencephalography (EEG), as a typical non-invasive biosensing signal, reflects individual emotional changes by recording the brain’s neural activity in response to various external stimuli. However, the significant differences in brain activity among individuals and the complex interrelationships between EEG channels notably hinder the accuracy of emotion decoding in non-invasive biosensing scenarios. To address this challenge, this paper proposes a two-discriminator domain adversarial neural network method (TD-DANN). The proposed method aims to obtain more generalized and individualized emotion feature representations through adversarial learning. Specifically, graph convolution is utilized to extract features from EEG signals. By modeling the EEG channels as graph nodes, the adjacency matrix can be dynamically learned to capture the complex relationships between different channels during emotion generation. Moreover, we design a domain discriminator and an individual discriminator. The domain discriminator is used to minimize the difference in feature distribution between the source and target domains. It is able to obtain discriminative features with universality. The individual discriminator is used to learn discriminative features consistent with the individual’s brain activity. It can enhance the adaptability to the individual’s emotion. The experimental results show that the TD-DANN achieves promising recognition accuracies of (98.45 ± 2.38)% and (89.45 ± 5.87)% for subject-dependent and subject-independent experiments on the SEED dataset, respectively. The proposed method attains recognition accuracies of (84.40 ± 8.70)% and (77.13 ± 7.97)% for subject-dependent and subject-independent experiments on the SEED-IV dataset, respectively. These results validate the effectiveness of the TD-DANN in the emotion decoding problem.

Hypergraphs are generalization of graphs, introduced by Berge. In an ordinary graph, an edge connects exactly two vertices, whereas in hypergraphs, a hyperedge can join any number of vertices. Hypergraphs have applications in the field of Computer Science, Machine learning, Neural networks etc . In this paper, we focus on the Knight’s hypergraph in which the squares of a chessboard are taken as vertices and each hyperedge include a vertex and all the vertices which are reachable by a knight in one move. We find the Adjacency matrix, Laplacian matrix, their eigenvalues and corresponding energies of the Knight’s hypergraph with the help of Python programming.

Abstract Kidney stones, formed by the accumulation of minerals and salts in the kidneys, can lead to severe pain and significant complications. Timely and accurate detection of these stones, particularly in medical imaging like CT scans, is crucial for effective treatment. This study introduces a novel method for detecting and classifying kidney stones in CT images, addressing the challenges posed by the variety of stone shapes and sizes, image noise, and the difficulty of identifying small or atypical stones. The proposed method utilizes a graph attention network to enhance detection accuracy and minimize identification errors. Initially, keypoints are extracted from the images using the ORB algorithm. These keypoints are considered as nodes in a graph, and using an appropriate threshold, the nodes are connected to each other to form the graph. The relationships between them are then defined through an adjacency matrix. Subsequently, for each node, a feature vector is obtained by using the pixel values within a 5 × 5 window centered on the corresponding node, and a feature matrix is constructed for the entire graph. The graph is then transformed into an embedded space, where features are updated through the graph attention network. This attention mechanism allows the model to discern the significance of each node and its connections, effectively extracting structural information. The updated embedded data is subsequently fed into a deep neural network for training and classification. Evaluation of the model’s performance is conducted using a dataset of coronal CT scan images. The results indicate that the proposed model achieves an impressive accuracy of 99.14% in detecting kidney stones, significantly outperforming existing methods, particularly in identifying small and atypical stones. This study highlights the effectiveness of integrating graphical features with attention mechanisms for enhanced medical image analysis, aiding in the early and accurate diagnosis of kidney stones.

This study developed a novel hybrid Graph Convolutional Network–Long Short-Term Memory (GCN–LSTM) model to forecast greenhouse gas (GHG) emissions across multiple country sectors, aiming to enhance climate policy. We analyzed 52 years (1970–2022) of GHG emissions data (CO₂, CH₄, N₂O, F-Gases) from 163 countries and eight sectors (Agriculture, Buildings, Fuel Exploitation, Industrial Combustion, Power Industry, Processes, Transport, Waste), sourced from the EDGAR v8 database. The GCN adjacency matrix captures spatial relationships on a weighted sum of Haversine distance and cosine similarity, while the LSTM models temporal dynamics. Data preprocessing includes min-max scaling and outlier handling with Interquartile Range capping. The model was trained on 70% of the data, validated on 15%, and tested on 15%, using Mean Squared Error (MSE) loss and the Adam optimizer. The performance was evaluated with Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Coefficient of Determination (R²). The GCN–LSTM model outperformed baseline models (ARIMA, Simple LSTM, Stacked LSTM), achieving the lowest MAE (0.0207 in Waste) and highest R² (0.9756 in Waste). Model interpretability highlighted strong regional connections, such as Thailand–Cambodia in the Waste sector, suggesting that spatial and temporal dependencies offer superior forecasting accuracy, informing targeted climate action.

Beyond structural bias: Improving circRNA-disease association prediction with Multi-Hop Neighborhood Hierarchical Fusion.

This paper introduces the Frank graph, a special weighted digraph with self-loops whose adjacency matrix corresponds to the Frank matrix, a special Max matrix. We derive bounds for the spectral radius and energy of the graph and illustrate these results with numerical examples. By investigating the spectral properties and extending the analysis of energy to weighted digraphs with self-loops, this study provides new insights and serves as a reference for future research on graphs with specialized adjacency matrices.

As skeletal data can be collected non-invasively while preserving patient privacy, it is widely used in public medical datasets to document patient behavior. Autism Spectrum Disorder (ASD) is characterized by significant behavioral heterogeneity, reflected in the topological structure and dynamic evolution of skeletal movements. This complexity poses substantial challenges for skeleton-based behavioral analysis. Existing methods struggle to effectively utilize behavioral evolution for subject-specific reasoning, leading to suboptimal representations that lack diagnostic relevance for autism. To address this limitation, we propose a Behavioral Evolution-based Edge Reconstruction (BER) Strategy for learning autism-related behavioral representations. By reconstructing a high-granularity adjacency matrix that spans both spatial and temporal dimensions, utilizing dynamic evolution and spatial location information, BERGCN enhances behavioral reasoning. Specifically, we first compute channel-level spatial and temporal edge reconstruction parameters by performing feature compression and targeted convolution operations on the differences between neighboring frames. Based on these, the spatial edge reconstruction module is designed by combining a generic attention map with two personalized attention maps, while the temporal edge reconstruction module is implemented using flexible frame replace ment and weighted aggregation. Finally, we investigate both single-modal and multimodal network architectures under various fusion strategies. We evaluate BERGCN on three autism clinical tasks and a benchmark action recognition dataset. Experimental results demonstrate competitive performance, showing improved sensitivity to subject-specific behavioral patterns while maintaining computational efficiency.

Accurate spatiotemporal forecasting underpins high-stakes decision making in smart urban systems, from traffic control and energy scheduling to environment monitoring. Yet two persistent gaps limit current models: (i) spatial modules are often biased toward low-pass smoothing and struggle to reconcile slow global trends with sharp local dynamics; and (ii) the graph structure required for forecasting is frequently latent, while learned graphs can be unstable when built from temporally derived node features alone. We propose SpeQNet, a query-enhanced spectral graph filtering framework that jointly strengthens node representations and graph construction while enabling frequency-selective spatial reasoning. SpeQNet injects global spatial context into temporal embeddings via lightweight learnable spatiotemporal queries, learns a task-oriented adaptive adjacency matrix, and refines node features with an enhanced ChebNetII-based spectral filtering block equipped with channel-wise recalibration and nonlinear refinement. Across twelve real-world benchmarks spanning traffic, electricity, solar power, and weather, SpeQNet achieves state-of-the-art performance and delivers consistent gains on large-scale graphs. Beyond accuracy, SpeQNet is interpretable and robust: the learned spectral operators exhibit a consistent band-stop-like frequency shaping behavior, and performance remains stable across a wide range of Chebyshev polynomial orders. These results suggest that query-enhanced spatiotemporal representation learning and adaptive spectral filtering form a complementary and effective foundation for effective spatiotemporal forecasting.

Abstract The one-way model of quantum computation is an alternative to the circuit model. A one-way computation is driven entirely by successive adaptive measurements of a pre-prepared entangled resource state. For each measurement, only one outcome is desired; hence a fundamental question is whether some intended measurement scheme can be performed in a robustly deterministic way. So-called flow structures witness robust determinism by providing instructions for correcting undesired outcomes. Pauli flow is one of the broadest of these structures and has been studied extensively. It is known how to find flow structures in polynomial time when they exist; nevertheless, their lengthy and complex definitions often hinder working with them. We simplify these definitions by providing a new algebraic formulation of Pauli flow. This involves defining two matrices arising from the adjacency matrix of the underlying graph: the flow-demand matrix M and the order-demand matrix N . We show that Pauli flow exists if and only if there is a right inverse C of M such that the product NC forms the adjacency matrix of a directed acyclic graph. From the newly defined algebraic formulation, we obtain O ( n 3 ) algorithms for finding Pauli flow, improving on the previous O ( n 4 ) bound for finding generalised flow, a weaker variant of flow, and O ( n 5 ) bound for finding Pauli flow. We also introduce a first lower bound for the Pauli flow-finding problem, by linking it to the matrix invertibility and multiplication problems over F 2 .

ABSTRACT Accurate traffic flow prediction is essential for mitigating road congestion. However, most existing methods use static or dynamic graphs to model spatial dependencies, often neglecting the dynamic relationships between road nodes that evolve over time, limiting their ability to capture complex spatiotemporal traffic patterns. To address this, we propose the Interactive Progressive Dynamic Spatiotemporal Graph Convolutional Network (STIPDG). The model constructs a progressive adjacency matrix by learning trend similarities between nodes, adapting to dynamic changes in traffic data. Based on this matrix, an interactive dynamic graph convolutional network synchronously captures spatiotemporal dependencies through an interactive learning strategy. Additionally, a progressive dynamic graph generation module accurately captures evolving spatial correlations within the traffic network. Experimental results show that STIPDG significantly outperforms baseline methods in prediction accuracy.

Decoding the neural basis of sensory phenotypes in autism.

Digital twins (DTs) have seen widespread application across industries, enabling deep integration of cyber–physical systems. However, previous research has largely focused on domain-specific DTs and lacks a universal, cross-industry modeling framework, resulting in high development costs and low reusability. To address these challenges, this study proposes a DT modeling method based on hierarchical decoupling and topological connections. First, the system is decomposed top–down into three levels—system, subsystem, and component—through hierarchical functional decoupling, reducing system complexity and supporting independent component development. Second, a method for constructing component-level DTs using standardized information sets is introduced, employing the JSON-LD language to uniformly describe and encapsulate component information. Finally, a topological connection mechanism abstracts the relationships between components into an adjacency matrix and assembles components and subsystems bottom–up using graph theory, ultimately forming the system-level DT. The effectiveness of the proposed method was validated using a typical surface water purification system as a case study, where the system was decomposed into four functional subsystems and 12 types of components. Experimental results demonstrate that the method efficiently enables automated integration of DTs from standardized components to subsystems and the complete system. Compared with conventional monolithic modeling approaches, it significantly reduces system complexity, supports efficient component development, and accelerates system integration. For example, when the number of components exceeds 300, the proposed method generates topology connections 44.69% faster than direct information set traversal. Consequently, this approach provides a novel and effective solution to the challenges of low reusability and limited generality in DT models, laying a theoretical foundation and offering technical support for establishing a universal cross-industry DT modeling framework.

A critical aspect of reliability analysis is selecting an appropriate structure-function to describe system behavior. Existing methods rely on identifying minimal path sets (MPSs) and minimal cut sets (MCSs) to model the reliability bounds of two-terminal systems. Accurate reliability bounds are crucial for analyzing critical system states. This article introduces a new algorithm that solves the sum of disjoint products (SDP) problem using a multivariable inversion technique to directly transform the structure-function into a simplified disjoint reliability expression. The proposed method combines two matrix-based algorithms that use the system graph's connection and adjacency matrices to efficiently find MPSs and cut MCSs. The primary objective is to determine accurate lower and upper reliability bounds for complex communication networks, including bridge and double-bridge types. The algorithms were implemented in Mathematica 13 and tested on 14 benchmark systems, demonstrating high accuracy and computational efficiency.

Multi-parameter prediction of seawater quality based on dynamic spatio-temporal relationship network.

The determinant of the adjacency matrix of a quaternion unit gain graph

A spatial-temporal fusion based nonlinear causality analysis framework for root cause diagnosis of faults in nonstationary industrial processes with asymmetric distribution.

Dynamic graph transformation with multi-task learning for enhanced spatio-temporal traffic prediction.

On the spectral radius of extended adjacency matrix of a digraph

GAT-Enhanced TabNet model with heterogeneous tabular and dependency graph information feature fusion for multi-disease coexistence risk prediction.

Brain functional network analysis models the brain as a graph of regions of interest (ROIs) and quantifies the correlations across different regions derived from functional magnetic resonance imaging (fMRI). Recently, artificial intelligence-based brain functional network analysis methods have demonstrated exceptional performance in diagnosing related neurological disorders. These approaches primarily focus on extracting relevant information from global connectivity patterns to analyze functional brain networks. However, medical research indicates that the impact of brain disorders predominantly manifests in localized functional connections among disease-relevant regions. Treating all connections equally risks introducing interference from irrelevant brain regions, thereby compromising diagnostic accuracy. To address this issue, we propose a novel sub-connection learning method that effectively identifies diagnostically specific connections while suppressing ineffective redundant connections. Specifically, we begin by employing a dynamic functional connectivity construction strategy to generate a functional connectivity matrix encapsulating dynamic features. Subsequently, we design a sub-connection Mask Learning strategy, which employs a multi-head self-attention mechanism to adaptively learn connection masks from functional connectivity matrices, enabling the capture of disease-specific connections and the suppression of noise connections. Additionally, we introduce a Multi-mask Fusion strategy and a Mask Iterative Optimization strategy to further enhance mask quality. Experimental results demonstrate that our model outperforms state-of-the-art methods on the ABIDE I and ADNI datasets, achieving accuracies (ACC) of 72.30% and 80.99%.