graph neural networks

Graph Neural Networks for Social Network Analysis

This paper explores the application of Graph Neural Networks (GNNs) to social network analysis. Traditional social network analysis methods often struggle to capture complex, hidden relationships and patterns within networks, particularly in large and dynamic social structures. GNNs offer a novel approach by directly learning representations of nodes based on their network context. This work details the core concept of representing social networks as graph structures and leveraging GNNs to perform tasks such as node classification and link prediction. We demonstrate the potential of GNNs to uncover valuable insights by modeling relationships as a graph, and learning from the network topology itself. The results suggest GNNs can provide a significantly improved analytical tool for understanding social networks compared to conventional methods. ---

High-Dimensional Data Structure Auto-Generation

This paper investigates the development of a novel data structure auto-generation system utilizing machine learning. Recognizing the challenges posed by high-dimensional data, this research proposes a method for automatically constructing data structures like matrices and graphs, thereby simplifying data processing and analysis. The system leverages neural networks to learn inherent data structure patterns and generates these structures with minimal human intervention. The core mechanism centers around the automated extraction of key structural features from high-dimensional data, enabling the generation of robust and well-structured data representations. This approach offers a significant advancement in data handling, addressing the limitations of traditional methods.

Based on Multi-Modal Collaborative Learning for Latent Variable Discovery

This paper investigates the application of multi-modal collaborative learning for latent variable discovery. Traditional methods for uncovering hidden variables predominantly focus on single modality data, often overlooking the synergistic information present when multiple modalities are available. We propose a novel framework leveraging multi-modal collaborative learning to enhance the accuracy and robustness of latent variable discovery. The core of our approach involves fusing data from diverse sources – such as visual, auditory, and textual information – and utilizing a graph neural network (GNN) based multi-modal learning framework. This framework maps each modality into a unified graph structure, learns inter-modal relationships using GNNs, and subsequently infers latent variables based on these learned associations. We demonstrate the effectiveness of this approach through a theoretical analysis and outline the key components and algorithms involved. The ultimate goal is to achieve more precise and reliable identification of underlying hidden factors within complex datasets.

Temporal Graph Neural Networks with Predictive Contextualization

This paper introduces Temporal Graph Neural Networks (TGNNs), a novel approach to graph neural networks designed to handle dynamic graphs where relationships evolve over time. Traditional Graph Neural Networks (GNNs) operate on static graphs, failing to capture the inherent temporal dependencies within systems. TGNNs address this limitation by integrating predictive contextualization. This involves utilizing learned predictions about future node states and edge relationships to augment the message passing process within the GNN layers. A recurrent component forecasts these future states, and attention mechanisms are employed to prioritize the influence of these predicted contexts during message aggregation. The core contribution lies in the unified framework combining graph dynamics with predictive learning, offering a significant improvement over static GNNs in applications involving evolving systems such as social networks, traffic networks, and biological systems. The proposed architecture enhances learning efficiency and improves predictive accuracy by considering the anticipated future state of the graph.

STOKES-DOT/GradTDDFT: GradTDDFT v1.1.0

GradTDDFT v1.1.0 GradTDDFT v1.1.0 provides a JAX-based framework for differentiable DFT, TDA/TDDFT, and neural exchange-correlation functional training. Highlights PySCF-aligned numerical grids: The standalone JAX grid now matches PySCF-style Lebedev quadrature, Treutler–Ahlrichs radial grids, NWChem pruning, and Becke partitioning for grid levels 0–9. Improved conventional DFT/TDDFT: Optimized restricted and unrestricted DFT, TDA, and full-TDDFT calculations for non-neural XC functionals, including matrix-free Davidson solvers and JAX GMRES for implicit SCF differentiation. Implicit excited-state differentiation: Supports implicit differentiation of TDA/TDDFT eigenvalues and eigenvectors without backpropagating through the full eigensolver history. Roadmap PySCFAD-based numerical integration — under testing Automatic-differentiation nonadiabatic coupling vectors Equivariant graph-neural-network atomic-orbital basis functions — under testing Dressed TDDFT beyond adiabatic linear-response TDDFT Full Changelog: https://github.com/STOKES-DOT/GradTDDFT/compare/v1.0.0...v1.1.0

Benchmarking predictive architectures for meningitis forecasting and outbreak detection using environmental drivers - code

This repository contains the analytical code for the manuscript, Benchmarking predictive architectures for meningitis forecasting and outbreak detection using environmental drivers. It provides the complete Python and PyTorch codebase to replicate our multi-architecture evaluation of early warning systems for epidemic meningitis across 40 countries within the WHO Regional Office for Africa (AFRO). Repository Contents: Spatiotemporal Deep Learning: PyTorch implementations of the Spatiotemporal Graph Neural Networks (ST-GNN) and the custom Zero-Inflated Negative Binomial Neural Network (ZINB-NN) designed for sparse case counts. Tabular Ensembles: Training and evaluation scripts for the AutoGluon gradient-boosted tabular models, including the two-stage hurdle classifier for binary outbreak detection. Feature Engineering: Spatial processing pipelines generating the deterministic gravity network from Overture Maps building footprints, alongside climate covariates like the binary Harmattan extreme flag. Evaluation Framework: Scripts calculating continuous magnitude estimation (R²) and binary alert detection (PR-AUC) across feature ablations.

Dynamic Semantic Embedding Network (DSE-Net)

This paper introduces the Dynamic Semantic Embedding Network (DSE-Net), a novel approach to understanding evolving data streams. The core claim is that by integrating semantic embeddings with dynamic graph neural networks, we can achieve continuous, context-aware understanding, overcoming the limitations of static embedding models. DSE-Net employs a multi-layered architecture: a Transformer encoder for initial semantic embedding generation, a dynamic graph neural network (GNN) to model relationships within the data stream, and a reinforcement learning (RL) module to optimize the GNN's structure and parameters adaptively. Crucially, the embedding itself is updated based on the GNN's output, creating an evolving semantic representation. This approach addresses the shortcomings of existing methods, which either rely on static embeddings or static GNNs, by providing a dynamic and learning system capable of adapting to changing semantic relationships. The key innovation lies in the synergistic combination of these techniques, leading to a more robust and nuanced understanding of dynamic data. ---

Dynamic Topological Semantic Learning (DTSL)

This paper introduces Dynamic Topological Semantic Learning (DTSL), a novel approach to semantic understanding based on interactive observation and dynamic topological graph construction. The core claim of DTSL is that a system can automatically learn and build semantic relationships between objects by observing and simulating their interactions, dynamically adjusting its topology to reflect these evolving relationships. DTSL leverages a combination of reinforcement learning and graph neural networks to achieve this. The system begins with an unstructured graph and utilizes observation units (e.g., visual sensors, text analyzers) to collect data on object interactions. A reinforcement learning agent then iteratively modifies the graph's connections – adding or removing edges – to maximize performance on specified tasks, such as relationship prediction or instruction execution. A graph neural network (GNN) represents and reasons about the graph structure, guiding the connection adjustments. Over time, DTSL constructs a dynamic topology that directly reflects the semantic relationships within the observed environment. This approach contrasts with traditional methods relying on pre-defined knowledge graphs or manually annotated semantic relations, offering a more adaptive and flexible solution.

Accelerating Mixed Discrete-Continuous Motion Planning via Neural Graphs of Convex Sets

Motion planning problems such as collision-free navigation and contact-rich manipulation can be naturally formulated as optimization problems that couple discrete decisions with continuous trajectories. The Graphs of Convex Sets (GCS) framework offers a practical solution to these problems. It represents discrete decisions as nodes of a graph and encodes continuous trajectories in the edges connecting them. However, the resulting optimization subproblems can become computationally prohibitive for online replanning. In this work, we propose a learning-based strategy to mitigate this limitation. Specifically, we replace the costly convex relaxation step required by nominal GCS with a single forward pass through a Graph Attention Network that predicts a set of highly probable candidate paths through the graph. A lightweight ranking network then orders these candidates by their estimated trajectory cost. Evaluating them in this order, we terminate our search early while still recovering a near-optimal motion plan. We validate the resulting pipeline across diverse robotic tasks, including collision-free motion planning for a 3D quadrotor and a 7-DoF manipulator, and planning through contact for planar pushing. Across both convex and non-convex cost and constraint settings, our approach yields up to two orders of magnitude speedup over nominal GCS while maintaining a 100% success rate, at the cost of some suboptimality in the recovered solutions. Code implementations and video demonstrations can be found at https://neural-gcs.github.io/.

TERE4AI: A Knowledge Graph-Based Tool for Generating EU AI Act Compliant Requirements

Machine Learning-Enhanced Tabu Search for Tactical Wireless Network Design

Designing high-performance tactical wireless networks under realistic operational constraints gives rise to challenging combinatorial optimization problems, where the evaluation of candidate solutions relies on detailed physical and traffic-aware models. Although classical metaheuristics such as Tabu Search offer effective mechanisms for exploring large search spaces, their computational cost remains high because numerous candidate moves must be evaluated at every iteration. In this paper, we propose a data-driven framework that improves the efficiency of Tabu Search by learning to guide its move selection process. Rather than altering the neighborhood structure, our approach exploits the information contained in the search trajectories generated during the optimization process. At each iteration, we record both improving and non-improving edge-based transformations together with a set of descriptive features capturing the structural, geometric, and performance characteristics of the network. This information is used to train a Graph Neural Network (GNN) that predicts the impact of candidate moves on the objective function. The trained model is then integrated into the Tabu Search algorithm to rank candidate transformations according to their predicted quality, thereby reducing the number of costly objective evaluations while maintaining an effective exploration of the search space. Experimental results on synthetic benchmark instances demonstrate that the proposed learning-assisted Tabu Search notably reduces computation time while consistently producing higher-quality solutions than the standard algorithm. These findings highlight the potential of combining machine learning with metaheuristics by leveraging the implicit knowledge embedded in search trajectories, paving the way for more efficient solution methods for large-scale network design problems.

Exploring multi-answer visual question answering with object detection: a systematic review

Visual question answering (VQA) is a challenging research area that enables machines to answer natural language questions based on visual content by jointly understanding images and text. Conventional VQA systems typically produce a single answer for each image–question pair. However, many real world visual questions are ambiguous or complex, allowing multiple valid answers to exist. This systematic literature review (SLR) focuses on multi answer VQA systems and the use of object detection, following the PRISMA 2020 guidelines. We analyzed 58 peer-reviewed journal articles retrieved from the Scopus database published between 2020 and 2025. Ten of these studies clearly stated that generating multiple answers was their main goal. Forty-eight others indirectly supported answer variability by using object-based or multi-instance reasoning. Through this review, we examine the current methodologies for supporting multi-answer generation, including model architecture, datasets, and evaluation metrics. Most multi answer generation approaches utilize attention mechanisms, graph neural networks, and transformer-based models. Additionally, we propose a taxonomy of multi-answer VQA organized along four dimensions. Limitations are identified in datasets and evaluation metrics (i.e., answer ambiguity/subjectivity). Future research should focus on improving model interpretability and designing an evaluation framework that incorporates subjective and context-sensitive responses.