graph neural networks

Messages Passed Along the Edges: A Interdisciplinary Mapping Review of Graph Neural Networks for Relational Data

This article presents a narrative review of Graph Neural Networks for Relational Data in the context of Artificial Intelligence. The literature on this topic has expanded substantially over recent decades, yet it remains fragmented across subfields, methods, and national research traditions. Drawing on an interpretive synthesis of representative contributions, the review reconstructs the historical development of the area, examines the conceptual foundations and definitional disputes that organize its debates, and maps the contemporary landscape of research, including the methodological shift toward data-intensive approaches and the institutional pressures that shape publication practice. Particular attention is given to the role of GNN and graph learning as organizing themes, and to the conditions under which findings from different research traditions can be brought into productive comparison. The review identifies three synthetic conclusions: the literature is cumulatively strong but organizationally weak; methodological pluralism is better understood as a resource than as a defect; and the growing practical salience of the topic raises the stakes of its unresolved conceptual questions. An agenda for future work is proposed, emphasizing integrative research designs, transparent synthesis practices, and the protection of definitional and infrastructural work on which cumulative progress depends. The article is intended as both a reference map for newcomers and a provocation for specialists in Artificial Intelligence.

AdaH-Chain: Deterministic capacity-aware GNNs sharding in heterogeneous blockchains

Abstract Modern blockchain sharding protocols increasingly deploy Graph Neural Networks (GNNs) to optimize state partitioning. However, executing these learning heuristics across heterogeneous networks introduces two funda-mental challenges. First, hardware-dependent floating-point arithmetic evaluates inconsistently across disparate instruction set architectures. This inconsistency triggers arithmetic state divergence and permanent consensus failures. Second, existing methods enforce uniform load distribution across shards. This ignores physical hardware disparities and forces powerful nodes to idle while waiting for resource-constrained validators, creating severe straggler bottlenecks. To address these challenges, we propose AdaH-Chain, a deterministic capacity-aware sharding framework. First, we design a Cross-ISA Deterministic Execution Engine. It replaces unsafe floating-point operations with a symmetric fixed-point quantization mechanism to mathematically guarantee bit-wise consensus safety. Second, we introduce a Capacity-Aware GNN Sharding algorithm. This mechanism maps extracted transaction subgraphs proportionally to verified hardware throughput limits, effectively eliminating the straggler effect. We evaluate AdaH-Chain using real transaction traces containing one million accounts from the Ethereum Mainnet. Experimental results confirm that the fixed-point engine preserves 94.7%of the original topological clustering accuracy while ensuring a 0% consensus divergence rate. By structurally aligning workloads with hardware capacities, AdaH-Chain delivers a peak throughput of 4,800 TPS, achieving a 1.9× gain over state-of-the-art learning protocols.

Consciousness Is a Phase Transition: A Two-Channel Thermodynamic Criterion and an Exclusion Principle

We propose that consciousness is a phase transition of a non-equilibrium dissipative structure, characterised by two independently controlled channels that must both be driven past threshold. The central result is an exclusion criterion: a macroscopic dissipative structure cannot be conscious unless it satisfies three necessary conditions - (i) an energy channel G > 1 maintaining non-equilibrium pumping, (ii) an information channel G_info(kappa) > 1 maintaining global phase coherence (with analytic critical point kappa_c = 1.8809), and (iii) topological closure via a self-referential triad (|SCC| >= 3). The framework traces a single causal chain: non-equilibrium driving -> Brusselator Hopf bifurcation -> sigmoid threshold -> phase-amplitude coupling (PAC) -> von Mises phase density -> Kuramoto network amplification -> G_info > 1 -> conscious phase transition. Each stage is derived from first principles (Appendices C-G). At the microscopic level, a basis-free trace-distance order parameter O(g) = 1/2 tr|rho+ - rho-| in two-qubit Gorini-Kossakowski-Sudarshan-Lindblad (GKSL) systems rigorously proves that the energy and information channels are independently controllable: a symmetrisation scan collapses O(g) from 0.306 to < 10^-4 as the absorption/emission asymmetry is removed, while a separate dephasing scan destroys coherence-carried asymmetry without affecting the energy flow. The topological closure condition is instantiated at multiple physical scales by a self-referential triad of three irreducible roles - reversible carrier, energy currency, irreversible anchor - which we identify in quantum coherence, aerobic metabolism, neural dynamics, and planetary geochemistry. We demonstrate the framework's explanatory power through clinical neuroscience and artificial intelligence. General anaesthesia abolishes consciousness by selectively collapsing kappa below threshold while metabolic pumping persists - a channel dissociation no single-channel theory predicts. Mindfulness meditation acts as a phase-locking mechanism that elevates kappa past kappa_c by suppressing Default Mode Network noise. Feedforward large language models fail all three conditions: their computational graph is acyclic, sustains no limit cycle, and operates as a closed system at inference. The widely observed model collapse under recursive self-training is the thermodynamic signature of the system relaxing toward equilibrium.

Quantum-Enhanced Graph Neural Networks

Quantum-Enhanced Graph Neural Networks (QEN) presents a novel approach to graph neural network (GNN) learning, leveraging the principles of quantum entanglement to enhance both the training process and the network's expressive capabilities. This research investigates a method for utilizing entanglement to accelerate training and improve the network's ability to represent complex relationships within graph data. We introduce a new architecture incorporating quantum entanglement-enhanced layer activations, exploring its potential to overcome limitations inherent in classical GNN training. The core mechanism focuses on utilizing entanglement to reduce computational complexity and improve the speed of learning. This work demonstrates promising results through theoretical analysis and preliminary experimental validation, suggesting a significant improvement in both training speed and model performance.

Life Operators: A Self-Evolving Framework for Multiscale Life Modelling

Medical AI is moving beyond recognition towards clinical dialogue and longitudinal prediction. Yet a central question remains: how would a patient’s state change under intervention? Statistical models learn future observations, whereas mechanistic models describe selected processes. Neither provides a common framework for representing patient state, coupling scales or revising failed assumptions. We propose Life Operators: task-bounded mappings that define three scientific roles. Perception operators infer task-relevant biological states from multimodal observations, Evolution operators propagate these states under natural or intervention-conditioned dynamics, and Generation operators map them to measurable signals. Each role may be realised by equations, statistical models, neural networks or hybrids. Bridge operators connect components with different variables, scales and time steps. Selected operators and bridges form task-specific Operator Graphs containing the smallest set of states and mechanisms sufficient for a declared claim. This modular structure also makes scientific revision localisable. An AI co-scientist may propose changes to states, operators, bridges or graph structure, while independent evidence determines which variants are retained, restricted or retired. Over time, validated components could accumulate into broader multiscale models of the human body and provide a computational foundation for medical artificial superintelligence.

Messages Passed Along the Edges: A Contemporary Synthesis Review of Graph Neural Networks for Relational Data

This article presents a narrative review of Graph Neural Networks for Relational Data in the context of Artificial Intelligence. The literature on this topic has expanded substantially over recent decades, yet it remains fragmented across subfields, methods, and national research traditions. Drawing on an interpretive synthesis of representative contributions, the review reconstructs the historical development of the area, examines the conceptual foundations and definitional disputes that organize its debates, and maps the contemporary landscape of research, including the methodological shift toward data-intensive approaches and the institutional pressures that shape publication practice. Particular attention is given to the role of GNN and graph learning as organizing themes, and to the conditions under which findings from different research traditions can be brought into productive comparison. The review identifies three synthetic conclusions: the literature is cumulatively strong but organizationally weak; methodological pluralism is better understood as a resource than as a defect; and the growing practical salience of the topic raises the stakes of its unresolved conceptual questions. An agenda for future work is proposed, emphasizing integrative research designs, transparent synthesis practices, and the protection of definitional and infrastructural work on which cumulative progress depends. The article is intended as both a reference map for newcomers and a provocation for specialists in Artificial Intelligence.

Graph Neural Network-Based Program Vulnerability Prediction

This paper investigates the application of Graph Neural Networks (GNNs) for program vulnerability prediction. Traditional methods for vulnerability detection often rely on static analysis and signature-based approaches, which can be limited in their ability to capture complex code relationships and identify novel vulnerabilities. This research proposes a novel framework that leverages the power of GNNs to learn structural information from program code, ultimately leading to improved vulnerability prediction accuracy. The core idea is to represent program code as a graph, where nodes represent code elements (e.g., statements, functions, variables) and edges represent relationships between them. GNNs are then trained on this graph to identify patterns indicative of vulnerabilities, such as insecure coding practices and potential attack vectors. The results demonstrate the effectiveness of this approach, showcasing the potential of GNNs to enhance vulnerability detection and risk assessment in software development. The key contributions of this work include the development of a graph-based representation for program code, the design of a tailored GNN architecture for vulnerability prediction, and the demonstration of improved prediction accuracy compared to existing methods.

Adaptive Topological Optimization via Graph Neural Networks

This paper introduces a novel optimization framework based on graph neural networks (GNNs) designed for geometric transformation optimization. Traditional optimization methods often rely on handcrafted objective functions and are limited in their ability to handle complex geometric transformations. This work leverages the power of GNNs to learn a mapping of geometric transformations, enabling automated optimization. We propose a method that utilizes a graph representation of the transformation space, where nodes represent geometric elements and edges represent the transformations applied to them. A GNN is trained to predict the optimal transformation sequence, allowing for efficient and robust optimization of complex geometric patterns. The core mechanism focuses on learning a robust representation of the transformation space through graph neural networks, facilitating the discovery of optimal geometric transformations. The paper demonstrates the effectiveness of this approach through comprehensive experiments on several challenging geometric transformation scenarios, highlighting its superior performance compared to traditional optimization techniques. The results underscore the potential of GNNs for automating geometric transformation optimization, particularly in scenarios involving intricate patterns and high-dimensional transformations.

Intelligent scheduling of aerospace precision forgings based on deep reinforcement learning and graph neural networks

Data-efficient exploration of atomic clusters via uncertainty quantification on complex potential energy surfaces

Machine learning interatomic potentials have become an effective method for exploring complex potential energy surfaces; however, their application to atomic clusters is frequently hindered by the high cost of sampling diverse isomer spaces and the difficulty in ensuring model generalizability across complex energy landscapes. While uncertainty quantification (UQ) offers a pathway to mitigate data scarcity, its efficacy in capturing continuous potential energy surface features and guiding active learning within the complex landscape of clusters remains systematically unverified. In this study, we developed and evaluated three different UQ frameworks based on advanced UQ methods and integrated with graph neural networks: Bayesian Neural Networks, Evidence Neural Networks (ENN), and Monte Carlo Dropout (MCD). We first validated these models on the MD17 dataset to establish baseline performance, followed by a rigorous assessment on complex cluster systems (Ta2N3- and LaSi24) to probe their decision-making mechanisms in high-dimensional spaces. Our results show that ENN effectively reflects data adequacy, while MCD exhibits excellent robustness, with an MSE of <0.3 eV2 in the first 20 epochs. Furthermore, active learning driven by MCD significantly reduces the computational overhead of first-principles calculations while maintaining high predictive accuracy. This study provides a physics-informed guideline for selecting UQ strategies, facilitating the autonomous and efficient discovery of stable cluster isomers.

Neural Network-Based Accelerated Model for Condensed Combustion Products in Solid Propellant

Graph Neural Network and Artificial Based Models in Preclinical Pharmacology: Predictivity, Opportunities and Challenges

Major challenges in preclinical pharmacology, such as low predictivity, high rates of attrition, prolonged development cycles, and high costs remain significant issues as they largely relate to how translational relevance of past experimental models. This work aims to assess the potential of artificial intelligence-based models, especially graph neural networks in enhancing the predictivity, opportunity, discovery and challenges solving in preclinical pharmacology. This is a synthesis review on machine learning, deep learning and graph-based architectures being used to predict molecular properties, toxicity, drug-target and drug-drug interaction, drug repurposing and de-novo molecule design and their methodological basis and application in preclinical workflows. The reviewed literature reveals that AI models, particularly graph neural networks, may be useful in learning complex chemical and biological relationships, leading to improved prediction of ADMET properties, toxicity endpoints, pharmacokinetics, and therapeutic efficacy, while reducing reliance on animal models and accelerating candidate prioritization. Despite these advances, significant limitations persist, including data scarcity and heterogeneity, limited interpretability, bias, scalability constraints, and challenges in generalization across biological systems and regulatory acceptance. In conclusion, artificial intelligence and graph neural network–based approaches represent a transformative paradigm for preclinical pharmacology by enhancing predictivity, efficiency, and translational relevance; however, their successful integration into drug development will depend on advances in high-quality data generation, explainable and hybrid modeling strategies, standardized validation practices, and ethical and regulatory alignment