Large-scale Complex Research Articles

Secure artificial intelligence at the edge.

Sensors for the perception of multimodal stimuli-ranging from the five senses humans possess and beyond-have reached an unprecedented level of sophistication and miniaturization, raising the prospect of making man-made large-scale complex systems that can rival nature a reality. Artificial intelligence (AI) at the edge aims to integrate such sensors with real-time cognitive abilities enabled by recent advances in AI. Such AI progress has only been achieved by using massive computing power which, however, would not be available in most distributed systems of interest. Nature has solved this problem by integrating computing, memory and sensing functionalities in the same hardware so that each part can learn its environment in real time and take local actions that lead to stable global functionalities. While this is a challenging task by itself, it would raise a new set of security challenges when implemented. As in nature, malicious agents can attack and commandeer the system to perform their own tasks. This article aims to define the types of systemic attacks that would emerge, and introduces a multiscale framework for combatting them. A primary thesis is that edge AI systems have to deal with unknown attack strategies that can only be countered in real time using low-touch adaptive learning systems.This article is part of the theme issue 'Emerging technologies for future secure computing platforms'.

Studying the geotechnical stability of the earth’s surface during the formation of different backfill mass types in quarry cavities

The research focuses on studying the geotechnical stability of backfill mass in the mined-out spaces of inactive quarries for restoring the earth’s surface in the conditions of the Kryvyi Rih region, where large-scale complex iron ore mining is conducted. Based on the analysis, it has been determined that in the region, to fill inactive quarry cavities and failure zones, the traditional method of filling them with dump waste rock is used. Given that the city of Kryvyi Rih is a densely populated, highly indutrialised agglomeration, the restoration of additional land areas could bring significant benefits for the economic development of the region, but the rock mass is not able to provide reliable geomechanical stability of the earth’s surface. The transformation of the physical state of the backfill mass from a loose to a monolithic state is proposed. In order to select alternative backfill methods, the quantitative and qualitative structure of the accumulated wastes of the mining-metallurgical complex that can be used as backfill materials is analyzed. Taking into account the significant volumes of accumulated beneficiation tailings and their limited utilization, it is recommended to use them as part of cemented paste backfilling, which will lead to an improvement in the environmental situation. The research methodology consists of laboratory tests of physical-mechanical properties of dump hard rocks, analysis of the properties of paste backfilling, and numerical modeling of stability of various types of backfill masses. It has been found that the mixture of rocks of 0...100 mm in terms of bulk density and voidness is close to the minimum fraction of 0...5 mm, which does not require the need to select a certain fractional composition. A «stress-strain» curve of a 0...100 mm rock mixture with a logarithmic relationship, characterized by three stages of strain, has been plotted. The strain modulus value has been calculated. The similarity of the mechanical characteristics of the studied mixture to the full-scale characteristics of dump hard rocks is substantiated. The modeling results show that the maximum strains in the case of rock backfilling reach 780 mm, while the calculations of paste backfilling show a significant decrease in the subsidence value, reaching only 43 mm. The difference in the values of subsidence is due to a significant difference in physical-mechanical properties. Based on studying the hardening conditions of paste backfill mixtures and temperature dynamics in the region, a seasonal approach and the formation of a combined backfill mass are proposed to maintain the reclamation rate. The research findings are valuable for the development of the construction direction of reclamation in the Kryvyi Rih region and the rational use of restored land areas.

Development and application of a LC-HRMS method for simultaneous determination of ten illicit antifungal drugs in herbal bacteriostatic products.

The widespread use of sub-therapeutic antibiotics may lead to treatment failures, increased resistance to antibiotics, and even the encouragement of the formation of superbugs. Fraudulent herbal medicines that actually contain unlabeled active ingredients are a global problem. Therefore, streamlined and accurate analytical techniques that can identify and quantify a variety of antimicrobials in suspected illegal products are required. In the present work, we developed a sensitive, robust, and accurate method that involves multi-step extraction, EMR-lipid-based pass-through cleanup, and UHPLC-Q/HRMS with an orbital ion trap instrument analysis for the simultaneous detection of ten illicit antifungal drugs (flucytosine, fluconazole, ketoconazole, naftifine hydrochloride, griseofulvin, bifonazole, econazole nitrate, elubiol, clotrimazole, and miconazole nitrate) in herbal disinfection products. This validated method was then applied to 210 authentic samples, resulting in 25% of the samples being positive for the drugs. The results suggested the need to strengthen controls in this field to detect illicit antifungal drugs in herbal bacteriostatic products, which represents a potential health risk for the consumer. On a global scale, more and more frequent routine monitoring is being carried out in this area. This study provides valuable insights into large-scale complex sample chemical profiling, and the method developed appears to be promising and applicable for high-throughput routine multiple antibiotic analysis in a complex sample matrix.

Development of a multidimensional centrality metric for ranking nodes in complex networks

Ranking nodes in complex networks is a fundamental task in network science, with significant applications in fields like social networks, bioinformatics, and information dissemination. Many existing methods struggle to efficiently capture the multifaceted relationships between nodes, particularly in large-scale networks. However, they often ignore the semantic relationships between nodes when gathering information. To address these challenges, this paper introduces SAASP, a Semi-local centrality measure that combines an Augmented graph with Average Shortest Path theory to efficiently identify influential nodes in complex networks. In SAASP, the augmented graph can represent global relationships between nodes as semantic similarities, allowing distant relationships to be considered in node ranking. Additionally, it incorporates an enhanced version of average shortest path theory to handle the computational complexity associated with large networks. By extracting local subgraphs for each node and considering the extended neighborhood concept, SAASP ensures scalability while preserving key structural properties. The integration of the augmented graph and average shortest path theory enables SAASP to accurately and efficiently identify influential nodes in large-scale complex networks with lower complexity. The effectiveness of the proposed metric is demonstrated through extensive experiments using the SIR (Susceptible-Infected-Removed) model and Kendall's coefficient, where it outperforms existing centrality measures on real-world datasets.

Key Node Identification Method Based on Multilayer Neighbor Node Gravity and Information Entropy.

In the field of complex network analysis, accurately identifying key nodes is crucial for understanding and controlling information propagation. Although several local centrality methods have been proposed, their accuracy may be compromised if interactions between nodes and their neighbors are not fully considered. To address this issue, this paper proposes a key node identification method based on multilayer neighbor node gravity and information entropy (MNNGE). The method works as follows: First, the relative gravity of the nodes is calculated based on their weights. Second, the direct gravity of the nodes is calculated by considering the attributes of neighboring nodes, thus capturing interactions within local triangular structures. Finally, the centrality of the nodes is obtained by aggregating the relative and direct gravity of multilayer neighbor nodes using information entropy. To validate the effectiveness of the MNNGE method, we conducted experiments on various real-world network datasets, using evaluation metrics such as the susceptible-infected-recovered (SIR) model, Kendall τ correlation coefficient, Jaccard similarity coefficient, monotonicity, and complementary cumulative distribution function. Our results demonstrate that MNNGE can identify key nodes more accurately than other methods, without requiring parameter settings, and is suitable for large-scale complex networks.

Constrained Dynamics, Stochastic Numerical Methods and the Modeling of Complex Systems

The workshop aimed to unite researchers from diverse fields of mathematics and statistics to explore the foundations of high-dimensional modeling and computational studies. It addressed recent advancements in numerical analysis, dynamical systems, and stochastic differential equations that support model reduction for large-scale complex systems.Incorporating targeted geometric structures, such as Riemannian manifolds, into large-scale statistical models is known to enhance the stability, reliability, and efficiency of numerical methods. However, algorithms are often presented in application contexts without adequate attention to their fundamental properties, limiting the adoption of these advanced modeling methods.The workshop emphasized understanding the fundamental properties of these structures, their impact on dynamics and stochastic dynamics, and the need to redesign algorithms to capture essential properties, aiming for robustness and suitability for high-performance computation.By bringing together numerical analysts, statisticians, and modelers, the workshop sought to improve the quality of methods and identify new model frameworks to guide future development.

Three-dimensional quasi-complete information numerical simulation of gravity anomalies and parallel computing on CPU-GPU

Abstract In order to improve the efficiency and accuracy of numerical simulation of gravity anomalies in large-scale complex models, this paper proposes a method for three-dimensional numerical simulation of gravity anomalies under arbitrary terrain, and implements its CPU-GPU parallel acceleration scheme. By performing a two-dimensional Fourier transform along the horizontal direction, three-dimensional partial differential satisfied by gravitational potential are transformed into one-dimensional ordinary differential equations in different wavenumbers, reducing computational and storage requirements. Moreover, the ordinary differential equations in different wavenumbers are independent of each other and have good parallelism. By retaining the vertical direction as the spatial domain and introducing traveling wave decomposition, the effects caused by upper and lower boundaries are eliminated. The vertical grid is arbitrary. A two-dimensional quasi-complete information Fourier transform is utilized to enhance both the accuracy and efficiency of the Fourier transform process, allowing both uniform and non-uniform flexible sampling. Based on the high parallelism, one-dimensional ordinary differential equations are solved in parallel using CPU, and quasi-complete information Fourier transform are computed in parallel using GPU. An abnormal sphere is designed to analyze the wavenumber spectral distribution characteristics of the anomalous potential and summarize the wavenumber sampling rules. The logarithmic uniform sampling in the wavenumber domain has high accuracy. Compared with Gauss-FFT, the quasi-complete information Fourier transform selects fewer wavenumbers and exhibits higher accuracy. The computational efficiency is greatly enhanced through the use of single-precision CPU-GPU parallel scheme, and a model with tens of millions of nodes only takes a few seconds. Two terrain models are designed to verify the algorithm’s adaptability to arbitrary complex terrains. This is significant for achieving fine inversion imaging and interpretation of gravity anomalies under large-scale complex conditions.

A trustworthy reinforcement learning framework for autonomous control of a large-scale complex heating system: Simulation and field implementation

Traditional control approaches heavily rely on hard-coded expert knowledge, complicating the development of optimal control solutions as system complexity increases. Deep Reinforcement Learning (DRL) offers a self-learning control solution, proving advantageous in scenarios where crafting expert-based solutions becomes intricate. This study investigates the potential of DRL for supervisory control in a unique and complex heating system within a large-scale university building. The DRL framework aims to minimize energy costs while ensuring occupant comfort. However, the trial-and-error learning approach of DRL raises concerns about the trustworthiness of executed actions, hindering practical implementation. To address this, the study incorporates action masking, enabling the integration of hard constraints into DRL to enhance user trust. Maskable Proximal Policy Optimization (MPPO) is evaluated alongside standard Proximal Policy Optimization (PPO) and Soft Actor–Critic (SAC). Simulation results reveal that MPPO achieves comparable energy savings (8% relative to the baseline control) with fewer comfort violations than other methods. Therefore, it is selected among the candidate algorithms and experimentally implemented in the university building over one week. Experimental findings demonstrate that MPPO reduces energy costs while maintaining occupant comfort, resulting in a 36% saving compared to a historical day with similar weather conditions. These results underscore the proactive decision-making capability of DRL, establishing its viability for autonomous control in complex energy systems.

Development of design configuration management system for large-scale complex system (LaDCoM)

Development of design configuration management system for large-scale complex system (LaDCoM)

Long-Term Passenger Flow Forecasting for Rail Transit Based on Complex Networks and Informer.

With the continuous growth of urbanization, passenger flow in urban rail transit systems is steadily increasing, making accurate long-term forecasting essential for optimizing operational scheduling and enhancing service quality. However, passenger flow forecasting becomes increasingly complex due to the intricate structure of rail transit networks and external factors such as seasonal variations. To address these challenges, this paper introduces an optimized Informer model for long-term forecasting that incorporates the influences of other stations based on complex network theory. Compared to the ARIMA, LSTM, and Transformer models, this optimized Informer model excels in processing large-scale complex transit data, particularly in terms of long-term forecasting accuracy and capturing network dependencies. The results demonstrate that this forecasting approach, which integrates complex network theory with the Informer model, significantly improves the accuracy and efficiency of long-term passenger flow predictions, providing robust decision support for urban rail transit planning and management.

Mass housing in transition: innovability in large-scale housing complexes

Mass housing in transition: innovability in large-scale housing complexes

Cascadable optical nonlinear activation function based on Ge-Si.

To augment the capabilities of optical computing, specialized nonlinear devices as optical activation functions are crucial for enhancing the complexity of optical neural networks. However, existing optical nonlinear activation function devices often encounter challenges in preparation, compatibility, and multi-layer cascading. Here, we propose a cascadable optical nonlinear activation function architecture based on Ge-Si structured devices. Leveraging dual-source modulation, this architecture achieves cascading and wavelength switching by compensating for loss. Experimental comparisons with traditional Ge-Si devices validate the cascading capability of the new architecture. We first verified the versatility of this activation function in a MNIST task, and then in a multi-layer optical dense neural network designed for complex gesture recognition classification, the proposed architecture improves accuracy by an average of 23% compared to a linear network and 15% compared to a network with a traditional activation function architecture. With its advantages of cascadability and high compatibility, this work underscores the potential of all-optical activation functions for large-scale optical neural network scaling and complex task handling.

Multiobjective design of automated order fulfillment systems: A case study for an existing surgical complex

We present a case study that introduces a systematic framework for designing an automated guided vehicle (AGV) order fulfillment system. The multi-stage framework integrated discrete event simulation (DES), super-efficiency data envelopment analysis (S-DEA) model, and Design of Experiments (DOE) methodologies for the design of an AGV system in an existing large-scale surgical complex within a comprehensive hospital in Singapore. Performance metrics derived from this model were consolidated into a single relative efficiency score using the S-DEA model. This efficiency score was then utilized in a DOE methodology, combined with a post-hoc model selection process, to robustly identify appropriate models for evaluating alternative system designs. The proposed framework optimizes AGV deployment by first identifying opportunities for AGVs to add value within the surgical system through a value-stream mapping approach. It then derives the optimal system design by balancing cost efficiency, manpower reallocation, and order fulfillment rates, facilitated by the S-DEA model augmented with the DOE approach and post-hoc model selection. Application of this framework to the surgical complex revealed several insights that can significantly enhance AGV efficiency prior to actual deployment. However, a key limitation of this case-study-based approach is its limited generalizability across different systems. Despite this, the methodology provides a robust foundation for AGV system design in similar healthcare environments.

Acoustic emission source location in complex structures based on artificial potential field-guided rapidly-exploring random tree* and genetic algorithm

Towards more accurate and easy-to-implement damage detection in large-scale complex structures, a novel acoustic emission (AE) source location method is developed based on artificial potential field-guided rapidly-exploring random tree* (APF-RRT*) and genetic algorithm (GA). APF-RRT*, which combines the excellent obstacle avoidance ability of RRT* with the path planning efficiency of APF, is introduced to adaptively estimate the shortest distances from the damage source to AE sensors. The shortest distances are obtained as the actual propagation distances of waves and then embedded into the modified error function, where GA is employed as an optimization scheme to evaluate the source location via iterations. Through the experiment on a full-scale high-strength bolt joint plate with a series of bolt holes, the effectiveness and superiority of the proposed method were validated. It achieved a better source location performance with lower mean absolute error and standard deviation than the time-of-arrival (TOA) method, delta-T mapping method, and machine learning-improved methods based on Gaussian process (GP) and artificial neural network (ANN), respectively. The primary contributions of the proposed method lay in abandoning the straight-wave-propagation assumption of the traditional TOA method by adaptively taking into account the geometric obstacles in complex structures, and removing the need for a large amount of training data and burdensome pencil lead break (PLB) tests required by data-driven location methods.

High-efficient sample point transform algorithm for large-scale complex optimization

Decomposition algorithms and surrogate model methods are frequently employed to address large-scale, intricate optimization challenges. However, the iterative resolution phase inherent to decomposition algorithms can potentially alter the background vector, leading to the repetitive evaluation of samples across disparate iteration cycles. This phenomenon significantly diminishes the computational efficiency of optimization. Accordingly, a novel approach, designated the Sample Point Transformation Algorithm (SPTA), is put forth in this paper as a means of enhancing efficiency through a process of mathematical deduction. The mathematical deduction reveals that the difference between sample points in each iteration loop is a simple function related to the inter-group dependent variables. Consequently, the SPTA method achieves the comprehensive transformation of the sample set by establishing a surrogate model of the difference between the sample sets of two cycles with a limited number of sample points, as opposed to conducting a substantial number of repeated samplings. This SPTA is employed to substitute the most time-consuming step of direct calculation in the classical optimization process. To validate the calculation efficiency, a series of numerical examples were conducted, demonstrating an improvement of approximately 75 % while maintaining optimal accuracy. This illustrates the advantage of the SPTA in addressing large-scale and complex optimization problems.

Traffic speed prediction of regional complex road network based on CapsNet and D-BiLSTM

Accurate and efficient short-term traffic prediction is of great significance in the study of regional traffic network. However, the complex and dynamic spatiotemporal correlation of traffic patterns makes the existing methods insufficient in learning traffic evolution in terms of structural depth and prediction scale. Therefore, this paper proposes a deep learning model combining capsule network (CapsNet) and deep bidirectional LSTM (D-BiLSTM). CapsNet is used to identify the spatial topological structure of the road network and extract spatial features, and D-BiLSTM network is integrated. The forward and backward dependencies of traffic states are considered at the same time, and the bidirectional temporal correlation of different historical periods is captured to predict the traffic of large-scale complex road networks in the target area. Experiments on real traffic network speed datasets show that the prediction accuracy of the proposed model is improved by more than 10% on average , which is better than other methods. It has high prediction accuracy and good robustness in traffic prediction of regional complex road networks.

A Bridge-based Algorithm for Simultaneous Primal and Dual Defects Compression on Topologically Quantum-error-corrected Circuits

Topological quantum error correction (TQEC) using the surface code is among the most promising techniques for fault-tolerant quantum circuits. The required resource of a TQEC circuit can be modeled as a space-time volume of a three-dimensional diagram by describing the defect movement along the time axis. For large-scale complex problems, it is crucial to minimize the space-time volume for a quantum algorithm with a reasonable physical qubit number and computation time. Previous work proposed an automated tool for bridge compression on a large-scale TQEC circuit. However, the existing automated bridge compression is only for dual defects and not for primal defects. This paper presents an algorithm to simultaneously perform bridge compression on primal and dual defects. In addition, the automatic compression algorithm performs initialization/measurement simplification and flipping to improve the compression. Compared with the state-of-the-art work, experimental results show that our proposed algorithm can averagely reduce space-time volumes by 53%.

ENHANCED PRUNING FOR DISTRIBUTED CLOSENESS CENTRALITY UNDER MULTI-PACKET MESSAGING

In recent years, identifying central nodes in large-scale complex networks has become a critical task, with applications ranging from social network analysis to traffic routing. While closeness centrality is a commonly used metric, its decentralized computation—especially in large networks—poses significant challenges due to high communication overhead. Previous approaches, such as pruning, aim to approximate centrality efficiently but often overlook the cost of exchanging multiple data packets in distributed settings. In this paper, we enhance the pruning method (a distributed approach for identifying central nodes using closeness centrality) by addressing this communication bottleneck. Specifically, we introduce a technique that leverages multi-packet messaging, reducing data loss and improving running times without compromising the accuracy of centrality estimates. Our approach also optimizes the communication load, leading to fewer exchanged messages overall. Empirical evaluation demonstrates that our method outperforms the original pruning technique in both message efficiency and computation time, while preserving the key properties of the baseline method. However, we also observe trade-offs in terms of increased memory usage and overhead, which we discuss in relation to network size and packet complexity. Our findings offer a more efficient solution for decentralized closeness centrality computation, presenting a promising direction for scalable network analysis.

K-plex-based community detection with graph neural networks

Community detection is an effective way to determine the structure and characteristics of a complex network. With the expansion of the network scale, traditional community detection approaches such as modularity-based optimization models face new challenges related to representing and learning the topological structure and node attributes from a large scale complex network. Moreover, in many practical applications, there is little knowledge about label information or the number of communities, which greatly limits the performance of existing supervised or semi-supervised community detection approaches. To solve these problems, in this paper, we propose a graph neural network-based unsupervised community detection approach, which first applies the k-plex to generate the community seeds, then uses a node sampling algorithm to reduce the network complexity, and finally constructs a graph neural network model to learn the relationships of the network nodes and assign the nodes to different communities. Extensive empirical studies on various scale networks demonstrate both the effectiveness and efficiency of the proposed approach. Our codes are available at https://github.com/lol12854/KPGN.

Detection of GPS Spoofing Attacks in UAVs Based on Adversarial Machine Learning Model.

Advancements in wireless communication and automation have revolutionized mobility systems, notably through autonomous vehicles and unmanned aerial vehicles (UAVs). UAV spatial coordinates, determined via Global Positioning System (GPS) signals, are susceptible to cyberattacks due to unencrypted and unauthenticated transmissions with GPS spoofing being a significant threat. To mitigate these vulnerabilities, intrusion detection systems (IDSs) for UAVs have been developed and enhanced using machine learning (ML) algorithms. However, Adversarial Machine Learning (AML) has introduced new risks by exploiting ML models. This study presents a UAV-IDS employing AML methodology to enhance the detection and classification of GPS spoofing attacks. The key contribution is the development of an AML detection model that significantly improves UAV system robustness and security. Our findings indicate that the model achieves a detection accuracy of 98%, demonstrating its effectiveness in managing large-scale datasets and complex tasks. This study emphasizes the importance of physical layer security for enhancing IDSs in UAVs by introducing a novel detection model centered on an adversarial training defense method and advanced deep learning techniques.