Accurate sub-pixel localization of chessboard corners plays a crucial role in enhancing stereo camera calibration and high-precision visual measurement. However, existing chessboard corner detection methods—whether traditional or learning-based—are susceptible to environmental conditions and image quality. Most also rely on complex post-processing techniques to reach sub-pixel accuracy, which not only increases computational burden but also limits robustness in challenging scenarios. To address these issues, this paper proposes a novel end-to-end method for sub-pixel corner detection. The approach incorporates a modified continuous heatmap loss to improve sub-pixel coordinate regression, along with a score loss that promotes prediction sparsity while suppressing false positive detections. Furthermore, we introduce a parallel detail compensation pathway to compensate for fine-grained spatial information lost during repeated down-sampling in the lightweight backbone network. This significantly improves the accuracy and robustness of corner localization. Experimental results demonstrate that our method achieves competitive sub-pixel localization performance on synthetic datasets and shows clear improvements in real-world stereo calibration and ranging tasks, confirming its practical value in vision-based measurement applications.

The article investigates the finite time synchronization (FTS) of variable-order fractional neural networks(VOFNN) with time delay under an event-triggered control (ETC) strategy. First, based on the definition of variable-order fractional derivative (VOFD), we propose a new lemma for VOFD inequality. Second, by designing a suitable ETC, we derive sufficient conditions for FTS between the drive-response systems using the Lyapunov function method. In addition, we accurately estimate the synchronization time and prove that the designed ETC avoids Zeno behavior. Finally, the theoretical results are verified through MATLAB simulations.

In software development project scheduling, the selection of work skills significantly affects both activity durations and salary levels, leading to variations in overall human resource use costs. Although extensive research exists on project scheduling and skill selection in software projects, most studies neglect the interactive impact of selection of work skills on both activity duration and salary level. Incorporating these interactions, however, complicates the problem by enlarging the solution space and increasing the complexity of parameter settings. This study proposes a mathematical model that integrates skill selection and project scheduling with the objective of minimizing total human resource use cost. A Quantum-behaved Particle Swarm Optimization (QPSO) algorithm is employed to optimize skill selections, enhanced by a cosine-function-based contraction–expansion coefficient to improve search performance. Furthermore, twelve priority rules are embedded into a serial schedule generation mechanism to derive feasible project schedules. Two groups of experiments are conducted to validate the effectiveness of the proposed approach and analyze the settings of key parameters. Results indicate that, compared to random assignment, standard QPSO, and linearly-controlled QPSO, the proposed algorithm achieves a lower human resource cost with slight impact on project duration. Moreover, human resource use costs initially decrease and then stabilize as the project deadline is extended, while they first decrease and then increase as the resource supply increases. Therefore, setting these two parameters at the earliest stabilization point and the minimum value point, respectively, can further optimize the cost objective.

Linear codes with few weights are widely used in various fields such as strongly regular graphs, secret sharing schemes, association schemes and authentication codes. The construction of linear codes with few weights and the determination of their weight distributions have always been an important research topic in error-correcting codes. In this paper, by selecting specific defining sets, this paper constructs two-weight, three-weight and four-weight linear codes with good parameters over finite fields with odd characteristic. By using some special exponential sums, the weight distributions of these linear codes are determined. The results generalize the corresponding results of Zhu and Xu (2017), Jian et al. (2019) and Zheng et al. (2021).

Computer numerical control machining is the cornerstone of modern manufacturing, a key challenge in this context is improving computational efficiency on free surface machining. This paper addresses this issue from two perspectives based on the geometric information of the surface. First, in terms of path planning, the concept of scallop height density is introduced, leading to the establishment of a scallop height metric. After reparameterization, using iso-parametric curves of the surface to be machined as tool paths enables the generation of uniform scallop height paths. Moreover, the parametric distance between paths can be used to estimate the scallop height error between adjacent paths. Regardince planning for parametric curves, existing algorithms have overlooked the geometric similarity between adjacent paths, which allows for motion similarity to be exploited. This paper proposes feedrate surface method, which enables the generation of feedrate curves for all paths by performing feedrate planning only on a subset of them. Compared to planning feedrate profiles for each path individually, this approach significantly enhances computational efficiency, making optimization-based feedrate planning methods have the opportunity for real-time applications. Furthermore, integrating these two aspects allows different machining processes such as finishing and semi-finishing to be consolidated into a single planning procedure. Experiments validate the effectiveness of the proposed theories and algorithms.

Accurate identification on the bond default mechanism will play a key supporting role in promoting the transformation of China’s bond market from scale expansion to quality improvement. Based on the credit bond default data from 2015 to 2023, this paper uses Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost) and other models to conduct empirical analysis, and selects the model with the best performance in predicting bond default. The results show that the CatBoost model performs well in predicting corporate bond defaults, and the visualization tool of Shapley Additive Explanations (SHAP) value effectively solves the ”black box” problem in machine learning prediction models and is able to visually highlight the impact of outliers. There is a correlation effect between supply chain concentration and customer concentration and credit default. Financial indicators maintain an irreplaceable importance in assessing the default risk of credit bonds, and are the key factors to measure the probability of default of credit bonds.

The optimization of loudspeaker arrays aims to select the optimal arrangement from multiple candidates of speaker positions, including determining the appropriate number and placement of speakers, which is crucial for improving sound field control. Focusing on the optimization of the loudspeaker array within a car cabin in consideration of sound field zone control, with the objective of maximizing the contrast in sound energy between bright and dark zones and enhancing spatial consistency, while minimizing sound field reconstruction errors and the number of speakers, this paper proposes a multi-objective optimization model to optimize the placement and quantity of speakers. Additionally, in view of the performance requirements in practical applications, a goal programming method is employed to select a speaker array configuration from the Pareto optimal solutions to meet the real-world needs. Simulations of the proposed multi-objective model were conducted at frequency points ranging from 0 Hz to 2000 Hz. Results demonstrate that the speaker array layout scheme obtained through the multi-objective optimization model can improve the performance indicators of sound field zone control.

This paper focuses on a class of composite optimization problems (COPs) of the form $\min_{x \in \mathbb{R}^{n}} \{ f(x) + \psi(x) \}$, where $f$ is a smooth convex function and $\psi$ is a possibly nonsmooth but proximally tractable convex function. To address the high computational cost associated with traditional second-order methods, we propose a gradient-regularized inexact Newton (IGRN) method, which extends the gradient-regularized Newton framework by incorporating approximate Hessian information. Specifically, our method replaces the exact Hessian matrix with an inexact approximation, such as one constructed via the limited-memory BFGS (L-BFGS) formula, under mild regularity conditions. We rigorously establish the global convergence properties of the proposed method: for general convex COPs, the IGRN method achieves a convergence rate of $O(T^{-1})$, and under strong convexity of $f$, it enjoys a global linear convergence rate. Furthermore, we develop an accelerated variant of the IGRN method by leveraging Nesterov acceleration technique. The resulting algorithm attains an improved convergence rate of $O(T^{-2})$ in the general convex setup. Numerical experiments on various logistic regression problems with different regularization terms and across multiple datasets are conducted. The results demonstrate that the proposed methods not only achieve competitive solution accuracy but also exhibit superior efficiency compared to several state-of-the-art algorithms for solving large-scale COPs.

Amidst global economic uncertainties and domestic structural transitions, bolstering corporate investment efficiency stands as a pivotal driver for China’s high-quality economic evolution. In April 2024, the Political Bureau of the CPC Central Committee first proposed nurturing “patient capital”— a capital form defined by long-term commitment, elevated risk tolerance, and strategic relational attributes, which aligns seamlessly with enterprises’ pursuit of sustainable, high-quality growth. This study empirically examines the impact of patient capital on corporate inefficient investment, utilizing a comprehensive dataset of Chinese A-share listed companies on the Shanghai and Shenzhen stock exchanges from 2008 to 2023. Our findings reveal that patient capital significantly mitigates both underinvestment and overinvestment, thereby enhancing investment efficiency. This positive effect is achieved through two key mechanisms: the resource mechanism, which facilitates the infusion of financial and business resources, and the governance mechanism, which strengthens corporate governance frameworks, internal controls, and external regulatory environments. Heterogeneity analysis uncovers that the effcacy of patient capital in curbing inefficient investment varies across firms with different technological capabilities, market competition intensities, and levels of fiscal support. By innovatively exploring patient capital from the perspective of corporate resource allocation, this research not only illuminates its critical role in optimizing investment efficiency but also provides actionable insights for refining capital market structures and enhancing capital allocation. These findings offer strategic guidance for policymakers and corporate decision-makers aiming to elevate investment efficiency and drive high-quality economic development.

Environmental governance, as a key element of global sustainable development, plays a crucial role in policy making, resource allocation, and achieving harmonious development between humans and nature. However, existing studies seldom addressed the impact of different types of environmental regulations on environmental governance costs, and overlooked the objectivity of indicator selection and the inherent information uncertainty of environmental governance data. To address these issues, this paper proposes a novel method for predicting environmental governance costs, and the main innovations include: (1) constructing a data-driven cost prediction model based on the extended belief rule base (EBRB); (2) incorporating the entropy weight method into the EBRB modeling process to ensure the objectivity of indicator selection; (3) integrating three types of environmental regulations into the EBRB-based cost prediction model; and (4) proposing an interval optimization model for EBRB to effectively resolve information uncertainty in environmental governance cost prediction. Based on empirical analysis of environmental governance data from 30 provinces in China, the results demonstrate that the proposed method can effectively reveal the differences in cost predictions under various types of environmental regulations and their interval variation characteristics, thereby providing a more scientific reference for environmental governance decision-making.

Accurate forecasting of carbon trading prices is crucial for market decision-making and risk management, yet existing methods face dual challenges: computational resource waste caused by cross-market repetitive modeling and insufficient forecasting accuracy in emerging markets due to data scarcity. To address the limitations of conventional transfer learning, including the imbalance between feature alignment and forecasting accuracy, as well as migration direction bias induced by fixed-weight mechanisms, this paper proposes a dynamic transfer optimization-based TimesNet model. First, a TimesNet network is pre-trained on source domain data to capture universal temporal features, addressing the deficiency of traditional models in cross-scale interaction modeling. Second, a dynamic transfer loss function is designed, incorporating a time-varying weighting mechanism to balance the cross-domain distribution loss and regression loss, thereby resolving the migration direction bias inherent in static strategies. Finally, model parameters are optimized via backpropagation to derive the TimesNet transfer forecasting model. Experimental results demonstrate that the proposed approach achieves lower error metrics compared to direct prediction in the target domain, exhibiting particularly strong robustness in data-scarce environments. For instance, when only 30% of the target domain data is used for fine-tuning, the transfer learning model shows notable performance advantages, with reductions in RMSE, MAE, and MAPE by 0.012, 0.011, and 0.25%, respectively. Moreover, compared to static transfer strategies, the dynamic mechanism effectively avoids interference from premature feature alignment on prediction tasks. Even with only 10% of target domain data for fine-tuning, the method consistently outperforms two static transfer strategies across RMSE, MAE, and MAPE, achieving reductions of 0.024/0.024, 0.024/0.021, and 10.27%/10.17%, respectively. To further evaluate the model performance, the DM tests were employed to verify forecasting reliability, and SHAP analysis was conducted to interpret model outputs. Results confirm that the forecasts align with realistic market behavior logic. This study establishes an interpretable transfer learning framework for carbon price forecasting, while expanding theoretical methodologies and practical tools for applying transfer learning to small-sample financial time-series scenarios.

Against the backdrop of global economic volatility and domestic growth transformation, scientifically assessing and mitigating the banking network systemic risk has emerged as a pivotal objective of macroprudential regulation. At present, the quantitative assessment of bank liquidation mechanism in systemic risk has not yet established a complete theoretical framework, especially the optimization analysis based on debt network structure. This paper innovatively proposes a maximal entropy model tailored to interbank debt networks under diverse liquidation mechanisms. Both theoretical and empirical analyses underscore the pivotal role of liquidation rules in restructuring debt connections, thereby substantially diminishing systemic risk exposure. Utilizing interbank transaction data from A-share listed banks from 2015 to 2024, we devise stress-testing scenarios for the four prevalent liquidation mechanisms, including proportional rule, priority rule, constrained equal awards rule, and constrained equal losses rule. Our investigation centers on the implications of these rules for network debt entropy and the survival rates of banks. The key empirical conclusions encompass: 1) The constrained equal awards rule exhibits the strongest robustness and capacity to maintain network diversity under various shock scenarios, with its risk-resilience advantage arising from dynamic regulation of excessive debt concentration. Its enhanced risk resistance arises from its capacity to dynamically curb excessive debt accumulation. 2) The priority rule, while enhancing individual survival rates, weakens anti-disturbance capabilities and exacerbates debt accumulation at core nodes under shocks, creating a "hotbed" for systemic risk. 3) The proportional rule holds certain robustness, but exhibits relatively limited risk mitigation capabilities under stress testing scenarios. 4) The constrained equal losses rule exposes risks of network fragmentation under extreme stress, significantly lowering system stability thresholds.

With the deep integration of information technology and manufacturing, the sharing economy has fostered an ecosystem model. Compared to traditional contract manufacturing arrangements, this model sees leading enterprises providing incubation support to start-ups, thereby influencing the profit distribution within the ecosystem.To elucidate the operational mechanisms of key elements, a supply chain comprising leading enterprises, start-ups and retailers was constructed. The study examined the impact of product brand strength, incubation capabilities and competitive intensity on supply chain profitability and competitive-cooperative strategies, analysing these factors from both the contract manufacturing model and ecosystem perspectives.Under the contract manufacturing model, collaboration between certain parties may to some extent undermine the interests of others. To a certain degree, the ecosystem model can effectively foster deep cooperation among supply chain members without compromising the interests of other stakeholders. Appropriate levels of competitive intensity can enhance product visibility and broaden the target audience. When a start-up's own brand gains consumer favour, consumer willingness to purchase products from leading brands also increases. At this juncture, leading enterprises will be more inclined towards the ecosystem model.This study provides theoretical underpinnings and practical guidance for the promotion of ecological chains, as well as for the optimisation of supply chain models and the formulation of collaborative strategies.

To better handle right-censored data and construct a survival analysis model with interpretability and predictive accuracy, this paper proposes the Deep Partially Linear Extended Hazard (DPLEH) model. The model’s key feature is its partially linear structure design, which allows specific covariates to be directly incorporated into the model as parametric components, thus enhancing the interpretability of the model results. Simultaneously, the model automatically learns nonlinear relationships through a deep learning framework, enabling it to capture nonlinear effects that are difficult to explicitly model in complex data structures and improve the predictive accuracy of the survival function. Furthermore, the interaction between covariates and survival time is introduced to eliminate the proportional hazards assumption restriction. This paper further analyzes the performance of various stochastic gradient descent algorithms in the model and proves the consistency of the survival function estimation under the model, providing a theoretical basis for its application. Through simulation studies, the proposed DPLEH model is found to perform well and outperform existing methods. Finally, the proposed model was applied to a dataset of primary breast cancer to validate its practical utility.

For the quadrotor unmanned aerial vehicle (QUAV) with prescribed performance, this paper proposes a predefined-time optimal trajectory tracking control strategy. Firstly, by employing error transformation and prescribed performance function, the system tracking error is constrained within a specified range. Secondly, an adaptive dynamic programming technique with an “actor-critic” structure is adopted to simultaneously optimize feedforward and feedback control signals. Furthermore, by using Lyapunov stability theory, the designed controller guarantees predefined-time stability of the closed-loop system while the tracking errors are constrained within a specified region and converge to a neighborhood of the origin within the predefined time. Finally, numerical simulation is conducted to verify the effectiveness and superiority of the proposed predefined-time optimal control algorithm.

In emergencies, panic emotions can easily trigger irrational behaviors, thereby amplifying systemic social risks. Addressing the issues of relatively disconnected emotion-behavior chains and idealized intervention mechanisms in existing models, this paper constructs a two-layer network coordinated stochastic control model that integrates emotion propagation, behavioral response, and government intervention. A nonlinear coupling structure is introduced between the emotion propagation layer and the behavior diffusion layer to capture the behavioral evolution of individuals driven by emotions, while coordinated stochastic control strategies simulate the effects and uncertainties of government intervention across multiple governance pathways. Control conditions for the convergence and stabilization of panic behavior are theoretically derived; simulations based on real-world cases validate the model’s fitting accuracy and control effectiveness. Results show that the synergistic mechanism of emotion guidance and government intervention, combined with flexible stochastic control, can effectively suppress panic behavior diffusion and improve the system recovery process, providing new modeling tools and theoretical support for behavioral governance in mass emergencies.

Under a trust mechanism, when a terminal within a dedicated P2P network initiates a resource transmission request to a cloud P2P system, the cloud peers must undergo identity verification against Certificate Revocation Lists (CRL) before being granted access to the dedicated system. This mechanism classifies resource-carrying peers in the cloud into two categories—Peer1 and Peer2—according to their historical record of resource transmission within the system. In this classification scheme, Peer2 with a transmission history are granted preemptive priority. Based on this, a Geom/Geom/$c$ queueing model is constructed in this study, and matrix geometric solution methods are employed to solve it, thereby obtaining quantitative indicators of the system's operational mechanisms and enabling an in-depth analysis of this P2P network system. Finally, regarding the preemptive priority characteristics of Peer2, a Stackelberg game equilibrium analysis is conducted on the model, yielding the unique arrival rates corresponding to the profit-maximizing conditions for both types of peers.

The application of AI-assisted detection in the medical field is becoming more extensive, effectively enhancing the diagnostic efficiency and accuracy of medical institutions. Capacity allocation can manage service rates and patient queues, thereby improving medical service quality. This study develops a two-stage capacity allocation model for medical systems using AI-assisted detection. It analyzes capacity allocation, profits, and social welfare in medical institutions with different characteristics. The findings indicate that medical institutions need to allocate base and safety capacity for detection and treatment stages by referring to the empowerment level of AI-assisted detection. This ensures base patient needs are met and provides flexibility to address demand fluctuations. While AI-assisted detection can boost the service rate in the detection stage, if medical institutions don't correspondingly increase treatment-stage capacity, it may cause system bottlenecks. Moreover, AI-assisted detection doesn't always enhance the profits and social welfare of medical institutions. This study offers theoretical support for the integration of AI technology and the medical field and provides guidance for capacity allocation decisions in medical institutions.

This paper considers an M/M/1 fluid-type queuing-inventory system with server's vacations, where the inventory items of the system are fluid-type items, and the system adopts a continuous review $(r,Q)$ replenishment strategy. Customers arrive at the system according to a Poisson process and take away a random size of items in inventory after the completion of their service. The server starts a vacation once the on-hand inventory is empty, and the vacation time obeys a general distribution. The arriving customers are not allowed to join the system if they find that the on-hand inventory is empty or the server is taking on vacations. Firstly, a modified system model with zero service time is considered and the steady-state probability distribution of the modified system is derived by using the supplementary variable method. Then, the product-form solution of the steady-state probability distribution of the original system model is obtained based on the steady-state probability distribution of the modified model. We further derive some performance indexes of the system and investigate the impact of the system parameters on the some performance indicators. Finally, the average cost function of the system is developed, and the optimal inventory policy and the optimal cost of the system are analyzed numerically.

The rapid evolution of social media has led to increasingly fragmented and stochastic information diffusion. In this context, organized public opinion manipulation through online troll factories has emerged as a significant challenge to cyberspace governance. Understanding the influence mechanisms of online troll factories is crucial for formulating effective countermeasures. Based on public opinion dynamics theory, this paper constructs a public opinion dynamics model for troll factory intervention in the social media environment. By introducing a random noise mechanism to capture the impact of stochastic information flows and establishing an asynchronous interaction model with local propagation features, we model troll factories as special agents with fixed opinions. Through quantitative analysis of the interaction dynamics between troll factories and ordinary individuals, we reveal the internal mechanism of how troll factories manipulate public opinion. Theoretical results show that group opinions can be synchronized in the mean sense to the preconceived opinion of the troll factory under the influence of an appropriate degree of stochastic information flow. This research deepens the understanding of troll factory influence, provides mathematical support for exploring their potential positive value in guiding public opinions, and offers a new theoretical perspective for the design of differentiated troll factory regulation strategies, thus contributing to the governance of the online public sphere

Promoting the construction of “waste-free cities” is an important practice to implement the twin goals of carbon peak and carbon neutrality, promote sustainable economic and social development, and build a beautiful China. This paper explores the impact of “waste-free cities” construction on corporate ESG performance using the 2019 “waste-free cities” construction pilot as a natural experiment. The study finds that the construction of “waste-free cities” has significantly improved the ESG performance of enterprises in pilot areas, and this phenomenon is more prominent in non-state-owned enterprises, manufacturing enterprises, and enterprises in regions with high environmental protection expenditures. Further results show that the construction of “waste-free cities” has increased the frequency of environmental terms in government work reports, the number of environmental penalties in the cities, and the number of green factories. At the enterprise level, it has alleviated managerial myopia and financing constraints, reduced the cost of equity capital, and increased credit access. Additionally, the enhancing effect of “waste-free cities” construction on ESG performance is more significant in enterprises that have not received environmental protection investments, environmental protection honors, or are in highly competitive industries. These results indicate that the construction of “waste-free cities” improves corporate ESG performance mainly through channels such as increasing government attention to the environment and enhancing stakeholder recognition of ESG. An analysis of its subsequent economic impacts reveals that the construction of “waste-free cities” can reduce the level of short-term debt used for long-term purposes by enterprises and drive the development of common prosperity of enterprises. This paper expands the research on “waste-free cities” construction and provides useful references for guiding enterprises to actively participate in social governance, making coordinated efforts to cut carbon emissions, reduce pollution, pursue green development, and boost economic growth and accelerating the green transition in all areas of economic and social development.

Climate change has become a key issue in global development, and actively addressing the challenges posed by global climate change has become a common consensus in society. In the context of big data, detecting and analyzing change points in meteorological data to assess the impact of policies or extreme events is particularly important. This article develops an estimation approach for generalized Logistic distribution change point model using generalized fiducial inference, with comprehensive comparisons to traditional maximum likelihood method and Bayesian method. Simulation results indicate that the point estimates and confidence intervals obtained based on generalized fiducial inference demonstrate good applicability. For change point estimation, the generalized fiducial point estimators based on mean show better effect, and the fiducial confidence intervals have more stable coverage rates within a reasonable range. Regarding scale and shape parameters, the generalized fiducial inference based on median is superior to other methods in most cases, and the fiducial confidence intervals based on highest posterior density can provide more accurate coverage, especially when estimating the shape parameter. Finally, the validity of the generalized fiducial inference in the generalized Logistic distribution change point model was verified using the monthly maximum rainfall data from the Beijing Meteorological Station.

In 2021, Wang gave a systematic study on the parity of coefficients of classical mock theta functions. Recently, a number of congruences modulo 4 for five mock theta functions have been proved by Chen and Garvan. Motivated by their works, we prove several new congruences modulo 4 and 8 for some Gordon and McIntosh's mock theta functions based on some identities involving mock theta functions and some results on congruences for the second-order mock theta function $A^{(2)}(q)$ due to Chen and Garvan. In particular, we confirm a conjecture on congruences modulo 8 for the second-order mock theta function $B^{(2)}(q)$.

The equivalence of multidimensional systems is closely related to the equivalence and reduction of multivariate polynomial matrices, and the Smith form of matrices plays a key role in matrix equivalence. The problem of the equivalence of multivariate polynomial matrices to Smith form is very complex and this paper studies the problem of equivalence between quasi weak linear multivariate polynomial matrices and their Smith form, characterizes the sufficient and necessary conditions for this type of matrix to be equivalent to its Smith form, and provides corresponding algorithms.

This paper studies the Smith form of polynomial matrices by using polynomial algebra isomorphism, discovers several new types of polynomial matrices, and characterizes the necessary and sufficient conditions for these new types of polynomial matrices to be equivalent to their Smith forms. These conditions can be judged by the theory and caculation of Gröbner bases,and there are corresponding algorithms.

In this paper, we consider a supply chain consisting of a manufacturer and an e-commerce platform. The manufacturer sells products through the platform's reselling channel and decides whether to encroach on the end market to gain additional profit. If the manufacturer chooses to encroach, it can do so via either direct encroachment or agency encroachment. The e-commerce platform, which possesses private demand information, decides whether to share this information with the manufacturer and simultaneously invests in promotional efforts to stimulate market demand .The results indicate that first, compared with direct encroachment, agency encroachment more effectively mitigates the double marginalization caused by wholesale pricing and leads to higher sales. Second, information sharing by the platform benefits the manufacturer; however, the platform is willing to share information only when the manufacturer adopts agency encroachment and the commission rate is high. Third, when promotional costs are high and commission rates are low, the manufacturer prefers agency encroachment, and information sharing further encourages this choice. Moreover, as demand volatility increases, both the manufacturer's preference for agency encroachment and the platform's willingness to share information first decline and then rise.

In this paper, the global leader-following consensus issue of discontinuous nonlinear fractional-order multi-agent systems is addressed via hybrid control strategy. Under the framework of fractional differential inclusion, one hybrid distributed control protocol with dynamic event-triggered mechanism is proposed. By utilizing Lyapunov stability theory and algebraic graph theory, the leader-following consensus for discontinuous fractional-order multi-agent system can be achieved, and Zeno behavior is excluded. Finally, a simulation example and a simulation comparison are respectively applied to verify the correctness of the obtained results and the effectiveness of the control strategy.

The intensive, efficient, green and low-carbon development path of new-type urbanization is highly consistent with new quality productivity theory. Based on panel data of prefecture-level cities in China from 2006 to 2021, new-type urbanization pilot policy is regarded as a quasi-natural experiment, and the impact of pilot policy on new quality productivity is estimated by using a multi-period difference in difference model. The study found that new-type urbanization pilot policy significantly promoted new quality productivity, and conclusion remained valid after a series of robustness tests, including parallel trend test, placebo test, replacement of explained variables, exclusion of other policy interference, propensity matching difference in difference model, and instrumental variable. Heterogeneity analysis shows that new urbanization pilot policy significantly promotes new quality productivity across different city regions, city sizes, city resource endowments, city economic development levels, city industrial structures, and city marketization levels. Specifically, it has a stronger promoting effect on new quality productivity in eastern cities, mega-cities and above, non-resource-based cities, cities with high economic development, cities with advanced industrial structures, and cities with a high degree of marketization. Further analysis shows that new-type urbanization pilot policy promotes new quality workers and new quality labor objects, and has a positive effect on new quality productivity through industrial structure rationalization and digital technology penetration. This study enriches the understanding of the policy effects of new urbanization and provides insights for accelerating the development of new quality productivity.

Multidimensional system reduction is often converted to the equivalence between multivariate polynomial matrices and their Smith forms. This paper focuses on the unimodular equivalence of a class of multivariate polynomial matrices to their Smith forms. A criterion for the unimodular equivalence of such matrices and their Smith forms is proposed, with the result extended to non-square and rank-deficient cases. Finally, a general algorithm for reducing such matrices to their Smith forms is presented and an example is provided to illustrate the algorithm.

With the rapid development of fresh product e-commerce, the quality of fresh product is widely concerned. For the supply chain of single fresh product supplier and single fresh product e-commerce platform, considering the fairness concern behavior of the platform, we construct the supplier-led Stackelberg game, the e-commerce platform-led Stackelberg game and the Nash equilibrium game model under the two scenarios of suppliers fulfilling and not fulfilling CSR, respectively, and study the impacts of fairness concern behavior, whether supplier should fulfill their responsibility and the differences in decision-making under different power structures of the platform through the solution of the six models, the comparison of profits and the numerical simulation. The study shows that: the platform's fairness concern behavior reduces the decision-making efficiency and negatively affects the supply chain's profit and social welfare; the supplier’s active fulfilling CSR can suppress the price and improve the quality of fresh product, but it will reduce their own profit and increase the platform's profit; under the power structure of Nash equilibrium, the supply chain system can achieve the optimal market performance, but the party that has the advantage of the power of the channel usually lacks the incentive to actively give up the power. Finally, this paper proposes a dual collaborative contract that achieves system optimality while achieving a Pareto improvement in the profits of supply chain members.

Remaining Useful Life (RUL) technology is of great significance to ensure the healthy operation of equipment. However, the current data-driven RUL research mainly focuses on large sample data, and there are few studies on small sample data. To solve this problem, this paper proposes a method based on feature engineering and model integration. First, perform feature engineering on the acquired product data, and construct new features from different perspectives; secondly, screen the features obtained by different feature construction methods to obtain the core features that can characterize the degradation information of aero-engines; finally, this paper selects appropriate base learners, and proposes a weighting method for model integration to obtain the final product RUL prediction model. Experiments are carried out on the C-MAPSS data set, and the results show that for the RUL prediction problem of small sample data, the method proposed in this paper improves the accuracy of product RUL prediction on the basis of effectively extracting product degradation features.

With the rapid development of e-commerce and logistics processing technologies, the cost of consumers’ searching for product information and purchasing has decreased substantially, leading to an explosive growth in the gray market. Motivated by profit-seeking incentives, some brand-authorized platforms may also be involved in gray market speculation activities. To effectively address the gray market speculation issues of authorized platforms, some manufacturers are contemplating the introduction of direct sales channels. Three scenarios are considered: the authorized platform engaging in no speculation, the platform engaging in speculation while the manufacturer does not introduce a direct sales channel, and the platform engaging in speculation while the manufacturer introduces a direct sales channel. A Stackelberg game model, with the manufacturer as the leader, is constructed to analyze these scenarios. The findings reveal: (1) When consumers in the low market (L) have a relatively low willingness to pay for the product, the platform tends to engage in speculation. As the willingness to pay increases among L market consumers, the manufacturer, without introducing a direct sales channel, will raise the wholesale price in the L market. With further increases in willingness to pay, even without any intervention from the manufacturer, the platform loses its incentive to speculate. (2) The introduction of a direct sales channel by the manufacturer can effectively boost its own profits and deter the platform's gray market speculation. (3) As consumers' willingness to pay for the manufacturer's direct sales products increases, the manufacturer's profits will rise while the platform's profits decrease. Therefore, manufacturers should strengthen their brand management for direct sales products and enhance consumer recognition.

This paper studies an M/M/1 queueing-inventory system with an emergency replenishment policy, where the inventory product is fluid type. There is one server in the system. After receiving service, the customer will take away a random amount of inventory products. When the inventory level reaches zero, the system will issue an order of emergency replenishment and the server will take a vacation. At the end of a vacation, if there is still no inventory in the system, the server will return to the system with probability $p$ or take another vacation with probability $(1-p)$ . It is assumed that the service time, the vacation time, the quantity of products required by customers, the lead times of the regular replenishment and the emergency replenishment all obey exponential distributions. Firstly, a modified system model with zero service time is considered, and the steady-state probability distribution is obtained by using Markov process theory and differential equation theory. Secondly, the steady-state probability distribution of the original system model is obtained by using the results of the modified system model, and it is found that the steady-state probability distribution has a product form. Based on this, we further obtain some performance measures of the system. Finally, the average cost function of the system is established. The optimal inventory strategy and the optimal cost of the system are analyzed by numerical examples.

This paper uses TVP-VAR model to study information spillover effects between international oil price structural shocks based on a new decomposition method and Chinese stock market sector. Furthermore, we employ GARCH-MIDAS model to examine the impact of geopolitical risk on spillover effects. The results indicate that: (1) Compared to the spillovers of volatility and tail risk, the spillover effects are mainly concentrated at the return level. (2) All industries are receivers of oil price shocks. The energy industry experiences the largest spillover effects from oil price shocks and is the largest receiver. (3) These information spillover effects are mainly driven by oil price risk shock. (4) The spillover effects exhibit time-varying characteristics. During sudden financial events, the connection between oil price shocks and Chinese stock market becomes more pronounced. (5) Geopolitical risk has a strong explanatory power for these information spillover effects. Against the backdrop of unprecedented geopolitical tensions, these findings not only help investors adjust their portfolios and reduce investment risks, but also provide valuable reference for policy-making and financial regulation.

This paper deals with a Tikhonov regularized second-order primal-dual dynamical system with slow vanishing damping for solving a linear equality constrained convex optimization problem. Under some mild conditions, we prove the trajectory of the dynamical system converges strongly to the minimal norm solution of the optimization problem. Additionally, we provide convergence rate results for the primal-dual gap, the objective residual, and the feasibility violation along the trajectory generated by the dynamical system. Furthermore, by conducting numerical experiments, we compare the obtained theoretical results with those reported in the existing literature.

In actuarial science and financial risk management, accurately characterizing the quantile function in the tail of a risk distribution is of critical importance for the measurement and control of extreme risks. Traditional risk measures, such as VaR, ES, CTE, and distorted risk measures, often exhibit large estimation biases and insufficient robustness under small-sample conditions or unknown distributions. To enhance the stability and consistency of risk estimation, this paper proposes a credibility-based estimation method for the quantile function, grounded in a Bayesian framework and the linear minimum mean squared error principle. The proposed method constructs an analytical credibility estimation model, effectively avoiding the computational complexity associated with high-dimensional posterior quantiles, and establishes a unified estimation framework applicable to multiple types of risk measures. Theoretical analysis demonstrates that the proposed estimator possesses desirable statistical properties, including conditional consistency, mean squared error convergence, and asymptotic normality. Numerical simulations further indicate that the method exhibits superior stability and accuracy compared to conventional empirical estimators in small-sample scenarios. Finally, we conducted an empirical analysis using daily data from six representative stocks in the Chinese stock market to evaluate the proposed estimation method. The results show that the method remains robust under conditions of high volatility and noise, adapts well to different market environments, and provides a reliable and practical approach for tail risk measurement.

Minimal linear codes play a crucial role in secure communications, particularly in secret sharing schemes and secure multi-party computation. Numerous studies have focused on constructing linear codes with few weights and, more importantly, minimality, using algebraic or geometric methods. In this paper, we propose two novel code constructions and employ Boolean functions and vectorial Boolean functions to construct several classes of binary linear codes with three and five weights. By applying the Ashikhmin-Barg theorem, we establish sufficient conditions for the minimality of these codes. Finally, we demonstrate the practical applications of the duals of these minimal linear codes in secret sharing schemes.

Information asymmetry between banks and enterprises and the lag in policy transmission constitute the core constraints limiting the effectiveness of technology innovation refinancing policies. This paper incorporates financing premium effects and credit rationing mechanisms into a DSGE model to investigate the impact of central bank digital currency (CBDC) on the effectiveness of technology innovation refinancing policies and analyze the primary transmission channels. The results demonstrate that implementing technology innovation lending through CBDC significantly enhances the optimization effects of technology innovation refinancing policies on both innovation input and output structures. Furthermore, mechanism analysis reveals that CBDC primarily enhances policy effectiveness by reducing firms' financing constraints and optimizing credit resource allocation. Particularly when CBDC accounts for more than 15% of technology innovation loans, the structural adjustment effects and welfare implications of technology innovation refinancing policies surpass those of direct monetary policy instruments. Extended research findings indicate that the "forward-looking conditional trigger mechanism" of CBDC contributes to further enhancing the policy's optimization effects on economic structure.

Advances in artificial intelligence have made generative rumors more challenging to distinguish from truthful information than conventional ones. Identifying the dissemination behavior of critical users accurately during the rumor propagation process has become a crucial strategy to mitigate large-scale rumor spreading. Given that the influence (weight) of rumor dissemination generally decreases as the propagation distance (number of hops) increases, this study assumes that the influence becomes negligible at three or more hops. Based on this assumption, we formulate the critical user identification problem as minimizing information interaction under a weighted two-hop constraint and examine mathematical properties of the objective function, such as non-convexity and submodularity. To solve this problem effectively, we propose a discrete gaining sharing knowledge based algorithm (DGSK), extending the original continuous framework to a discrete solution space. The algorithm incorporates novel strategies for population initialization, individual learning, and population update to enhance its performance. We also analyze the time complexity of DGSK and evaluate it on ten synthetic and real-world datasets. Comparative experiments with four mainstream heuristic algorithms across five evaluation metrics demonstrate that DGSK achieves superior performance in identification accuracy, stability, and convergence speed.

To address the issue of improving educational quality, this study deeply integrates the WSR (Physics-Affairs-Human Relations) systems methodology with system dynamics, and constructs a dynamic governance model for the continuous improvement of educational quality based on the analysis of causal relationships among various elements of the system. Simulation results indicate that the quality monitoring and evaluation system——which conducts assessments based on real data of talent cultivation quality (i.e., student development) and is linked to the reform of personnel and distribution mechanisms——serves as a value-oriented carrier functioning in feedback control. This system is not only an important means to optimize the structure and functions of the overall system but also a driving force for its operation. By driving comprehensive reforms in the system across three dimensions——namely, guarantee reform in the Physics (W) dimension, efficiency reform in the Affairs (S) dimension, and motivation reform in the Human Relations (R) dimension——it promotes cross-cycle iterative optimization and continuous improvement of the system, thereby achieving the enhancement of educational quality. The innovations of this study are reflected in the following three aspects: First, it organically integrates systems science with organizational behavior, providing a research perspective rooted in systems science and a research entry point based on organizational behavior for the field of educational quality management. Second, it breaks through the limitations of traditional static analysis and offers a logical analysis framework that combines static and dynamic approaches for this field; this framework is built on the integration of the WSR systems methodology and system dynamics, thus compensating for the lack of such logical analysis frameworks in the field. Third, it integrates cybernetics, system management theory, dissipative structure theory, and chaos theory to reveal the operational mechanism of the quality monitoring and evaluation system. This study embodies the academic value orientation of systems engineering——i.e., “practice-oriented and problem-driven”——and the perspective that “integration equals innovation”. Furthermore, this study provides decision-making suggestions and management insights for the effective advancement of policies regarding the reform of personnel and distribution mechanisms in the educational quality management system, the formation of a collaborative force for talent cultivation among local governments, schools, and society, and the improvement of educational quality centered on student development. Consequently