Constructing appropriate internal governance mechanisms to enhance organizational resilience is an important approach for companies to cope with uncertain shocks. This paper divides internal governance mechanisms into three levels: “shareholders, board of directors and incentives.” Using dynamic qualitative comparative analysis (QCA) and panel data from 64 AI A-share listed companies from 2018 to 2022, this study explores the impact of the internal governance mechanism configurations on organizational resilience under time effects. The findings indicate that the internal governance mechanisms need to be synergistic in order to effectively improve organizational resilience. Four internal governance mechanism configurations for companies with high organizational resilience can be summarized as technological, supervisory, and centralized models, with high compensation incentives being a common factor across all configurations. These configurations show time effects, and internal governance mechanisms need to be dynamically adjusted to enhance organizational resilience. This study enriches the understanding of factors affecting organizational resilience and deepens the research on internal governance mechanisms from the “governance bundle” perspective, providing practical insights for the management of AI companies.

Empirical studies on the influencing factors of false information verification behavior on social media have shown inconsistencies. This study systematically integrates relevant empirical research from both domestic and international sources to identify the key factors influencing users' false information verification behavior, providing insights for false information governance. Using a meta-analysis approach,the study extracts 98 effect sizes, a total sample size of 66 482, and 17 influencing factors from 35 empirical studies.Publication bias tests, heterogeneity tests, overall effect tests, and moderator effect analyses were conducted. The findings indicate that information-related, cognitive, emotional, and individual factors all exert varying degrees of influence on false information verification behavior. Specifically, information relevance, information importance, and information literacy have a significant impact on users' verification behavior. Additionally, social background, gender, discipline type, and education level moderate the relationships between related variables and false information verification behavior.

Precise governance is the guiding objective of cyber ideological security governance, and the powerful empowerment of big data provides both technological support and opportunities for the era of precise governance in cyber ideological security. There is an inherent alignment between big data and the precise governance of cyber ideological security, characterized by shared subjectivity, technological connectivity, and content interoperability, which forms the basis for such governance. However, challenges still exist in the process of big data empowerment for precise governance, including the absence of diverse governance subjects, the broadening of digital risks in governance, and insufficient understanding of the governance concept. To address these challenges, it is necessary to focus on three key aspects: innovating new paradigms for multi-party participation in governance, exploring new paths for risk warning and governance, and establishing a comprehensive strategic governance concept. These efforts will work synergistically to promote the development and construction of ideological security in cyberspace.

Against the backdrop of new engineering education, this study focuses on the alignment between university big data talent cultivation and enterprise demands. The aim is to reveal the matching degree between academic training models and industry needs, providing theoretical support and practical guidance for optimizing talent development systems. First, a co-occurrence network of academic papers and patent technologies is constructed using social network analysis to explore the relationship between academic research and enterprise technological needs. Second, a topic mining model based on Word2Vec-K-means-LDA is employed to extract core themes across different time periods. Finally, cosine similarity is used to measure the semantic alignment between topics. The results indicate that big data technology follows an evolutionary path from theoretical research to industrial applications, with strong alignment in areas such as knowledge graphs and data mining. However, university curricula still exhibit shortcomings in interdisciplinary integration and practical skill development, making it challenging to fully meet the demand for interdisciplinary talent in enterprises. In response, this study proposes strategies such as constructing interdisciplinary training systems and deepening university-enterprise collaboration to enhance talent development effectiveness and support the high-quality development of the big data industry.

Utilizing a dataset encompassing State-Owned Enterprises(SOEs) listed on the Shanghai and Shenzhen stock exchanges from 2010 to 2021, this study explores the impact of the Party Committee's participation in corporate governance on firm value and analyzes the influence and mechanism of the executive team's information fragmentation. The study finds that the degree of the Party Committee's involvement in corporate governance has a positive effect on firm value. Furthermore, when the Party Committee participates in the company's governance by entering the board of directors, the supervisory board, and the executive layer, it significantly affects firm value. The executive team's information fragmentation exhibits an inverse U-shaped moderating effect between the degree of involvement and firm value. When the intensity of the executive team's information fragmentation is low or high, it suppresses the positive impact of the Party Committee's participation in corporate governance on firm value. Only when the intensity is moderate can the Party Committee's involvement significantly enhance firm value.

Smart manufacturing not only fosters industrial upgrading at the macro level but also supports the sustainable development of enterprises at the micro level. This study empirically investigates the relationship between smart manufacturing and corporate “greenwashing” using data from A-share listed manufacturing firms in China from 2013 to 2022. The findings indicate that smart manufacturing plays a significant role in reducing corporate “greenwashing,” with this result remaining robust across various tests. Additional analysis through a quasi-natural experiment, adopting the smart manufacturing pilot policy (difference-in-differences, DID), further corroborates the robustness of these findings. Mechanism tests reveal that smart manufacturing mitigates “greenwashing” by reducing information asymmetry, easing financing constraints, and fostering green innovation. Heterogeneity analysis shows that the impact of smart manufacturing in curbing “greenwashing” is particularly pronounced in firms that voluntarily disclose environmental information, in heavily polluting industries, in companies located in inland cities, and in those with executives having international experience.

Autonomous driving is a complex and disruptive technology, and public acceptance plays a significant role in its application and development. This study explores the acceptance of autonomous vehicles among young people through a mixed analysis of big data and statistical models. The research finds that young people's attitudes toward autonomous vehicles are mixed, with slightly more negative than positive attitudes. On one hand, young people recognize the safety, efficiency, and comfort brought by the technology; on the other hand, they are concerned about risks such as data leakage and system security. Perceived risk, perceived ease of use, perceived usefulness, and trust all significantly affect young people's acceptance of autonomous vehicles, with trust serving as a mediating factor. Compared to older generations, young people have a more positive attitude toward autonomous vehicles and are more willing to embrace the technology, though they are also more concerned about risks. The attitudes of young people toward autonomous vehicle technology are neither blindly optimistic nor fearfully pessimistic, but rather objectively skeptical. In the future development of autonomous vehicles, it is necessary to reduce risks at the software, hardware, environmental, and data levels to improve public trust, while addressing the issue of responsibility allocation through “meaningful human control.”

In the context of China's new era, college students' values exhibit a range of complex and sometimes contradictory features: political beliefs marked by both alignment and tension; moral perspectives shaped by a mix of dedication and confusion; life attitudes characterized by both striving and passivity; learning motivations balancing idealism with utilitarianism; and interpersonal interactions that blend real and virtual dimensions. These characteristics are closely tied to the broader conditions of the new era, including the interplay of domestic and global changes (“two overarching contexts”), cultural diversification, the digitalization of lifestyles, and the widespread expansion of higher education. To enhance the effectiveness and relevance of values education for college students, it is essential to tailor guidance strategies to their specific value orientations and evolving socio-temporal contexts. This can be achieved through theoretical clarification, curriculum integration, platform development, supportive environments, and need-based approaches.

The rapid advancement of digital technology has introduced new challenges to anti-fraud education in universities. As a new paradigm of public administration and governance, collaborative governance is increasingly being applied as a reference framework in the modernization of university education. However, current anti-fraud education in universities faces several issues, including the absence of governance entities, insufficient objective linkage mechanisms and subjective domain awareness, and limited effectiveness of educational campaigns. University-based anti-fraud education operates within a stable and open system where governance resources, anti-fraud information, and co-governance dynamics circulate among multiple actors. The willingness of these actors to engage in collaborative governance is shaped by a four-dimensional structure of subjective driving forces, ultimately aiming to enhance students' “three senses” (security, participation, and gain), build safe campuses, and contribute to broader societal governance. By incorporating collaborative governance theory into anti-fraud education, universities can establish a multi-actor interactive governance network, explore effective points of integration for coordinated anti-fraud initiatives, and achieve a dynamic integration of traditional resources, outcome feedback, and accumulated experiences. This approach fosters a unified awareness of the anti-fraud education community in universities. Ultimately, a comprehensive and multi-layered governance network for anti-fraud education should be constructed, founded on “actor collaboration,” driven by “mechanism collaboration,” secured through “resource collaboration,” and oriented toward “awareness collaboration.”

Digital technologies have transformed the elements, processes, and forms of participation in traditional social innovation, making digital social innovation a popular topic in academic research. However, its disciplinary nature and conceptual definition remain unclear, resulting in the lack of a unified paradigm for digital social innovation research both domestically and internationally. By reviewing existing literature, this paper provides a comprehensive overview of the connotations, structural dimensions, and practical mechanisms of digital social innovation. Based on this, an analytical framework for digital social innovation is constructed, aiming to clarify and structure the research in this field and explore a new research paradigm for digital social innovation in the Chinese context.

As a soft institutional environment, can an optimal business environment enhance firm resilience to uncertainty shocks, serving as fertile ground for nurturing their adaptability? This study explores the impact of the business environment on firm resilience and the underlying mechanisms, focusing on the COVID-19 shock situation. Findings indicate an inverted U-shaped relationship between the business environment and firm resilience, where firms' risk-taking propensity positively moderates this relationship. Mechanism testing, based on a “resource-relationship-capability” framework, demonstrates that the business environment enhances firm resilience by optimizing resource allocation, improving government-enterprise relations, and stimulating innovation capacity. Heterogeneity analysis further indicates that the effect of the business environment on firm resilience varies significantly across different levels of regional development, industry factor intensity, and firm-specific characteristics (ownership structure, lifecycle stage, and scale). This study offers theoretical insights and practical recommendations to help enterprises fully leverage policy dividends from an optimized business environment to enhance firm resilience. It also guides government policymakers in designing more balanced and effective policies to improve the business environment.

This work presents 2-D numerical simulation results of Single-Event Burnout (SEB) for hardened 1.7 kV 4H-silicon carbide (SiC) power VDMOSFET. Research results prove that, compared with the conventional VDMOSFET with five-buffer layers (FB-VDMOSFET), the modified VDMOSFET has achieved better SEB performance under high Liner Energy Transfer (LET) value range. After reinforcement, the on-resistance of the VDMOSFET is lower under the action of the current spreading layer compared to the traditional structure. In addition, the integrated PN junction in the JFET region reduces the peak electric field under breakdown voltage from 3.66 MV/cm to 2.94 MV/cm. When the drain-source voltage (V_DS) is 1400V and the Linear Energy Transfer (LET) value is 0.5 pC/μm, the FB-VDMOSFET has burned out, while the improved VDMOSFET, due to the stepped source metal electrode contact and integrated PN junction, reduces collision ionization, lowering the maximum lattice temperature to 2 070 K. This reinforced structure can be used in the aerospace industry.

Driven by new-quality productive forces, the value connotation of the “Two Mountains” concept is continuously expanding. Centered on technological innovation and characterized by data, intelligence, and green development, new-quality productive forces promote the synergy of ecological industrialization, digital transformation, and sustainable green growth. This transformation not only broadens the value subjects of the “Two Mountains” concept by integrating technology enterprises, social organizations, and the public into the ecological economy but also enriches its value objects by facilitating the capitalization of ecological resources, the explicit recognition of ecosystem services, and the assetization of data. Meanwhile, new-quality productive forces optimize value measurement by making ecological benefits a key factor in economic decision-making, thereby fostering high-quality economic growth, sustainable social progress, and advancing ecological civilization to a higher level. Based on this value expansion, this study proposes practical pathways, including leveraging technological innovation to integrate ecological and economic development, promoting industrial upgrading to enhance the value of ecological resources, and cultivating ecological consciousness through social and cultural development. By deepening the value logic and practical model of the “Two Mountains” concept from the perspectives of human-nature, human-society, and self-environment relationships, this research contributes to the modernization of ecological civilization and provides an innovative perspective on Xi Jinping's ecological civilization thought.

In order to solve the problem of low matching accuracy and redundant feature points of SURF algorithm under complex environment changes, the SURF algorithm is optimized by combining the quadtree method for eliminating redundant feature points and the improved RANSAC algorithm for quickly obtaining the homography matrix. Firstly, the feature points generated by SURF algorithm are segmented by the quadtree method, redundant feature points are eliminated according to the maximum threshold of division depth and number of feature points, and rough matching is performed. Then, the feature points are sorted according to the distance between feature vectors, and RANSAC iteration is performed within the limited range, so as to obtain the optimized image matching result.Experimental results show that the average precision rate of the improved SURF algorithm is 96.47% and the average consumption time is 649.33 ms under a certain degree of rotation, zoom, blur, and illumination intensity variation. The improved SURF algorithm has good robustness, accuracy, and timeliness under complex environment changes.

Electric and magnetic fields play a crucial role in enhancing and optimizing the performance of Hall thrusters. Due to the geometric asymmetry of Hall thrusters, this study utilizes the COMSOL Multi-physics simulation platform for the construction of an accurate three-dimensional model, aiming to explore the characteristics of the electromagnetic field distribution within the thruster's channel. Furthermore, a qualitative analysis has been conducted to assess the impact of this field distribution on the propulsion performance, specifically in terms of thrust generation and efficiency.The results show that the radial magnetic field has a high-density region near the channel's throat,which exhibits a capturing effect on electrons emitted from the cathode. The axial electric field accelerates the ionized ions, providing propulsion for the thruster. These research findings offer theoretical guidance for both the improvement and the optimization of Hall thruster systems.

This paper introduces the problem that a trained fault diagnosis model for a power grid cannot be used when the grid topology is changed. In order to address this issue, a method based on convolutional neural network (CNN) and transfer learning is proposed. The model uses the voltage and current phase data collected by the synchronous phasor measurement unit (PMU) for the training of CNN. After the training process, the CNN model can accurately locate faults in the grid and determine fault types when the PMU equipment is limited. When the topology of the grid is changed, a new fault diagnosis model can be quickly trained by model-based transfer learning and a small number of fault samples under the new grid will be required. Moreover, the performance of the model is as good as the model trained by enough samples. The proposed method reduces the need of new samples in the training process of an untrained CNN model after the topology of the grid is changed.

To develop a band-pass filter with a favorable shape factor (SF), based on the substrate integrated waveguide (SIW) structure, a method is proposed that involves etching two asymmetric circular complementary split-ring resonators (CSRRs) on the top metal layer of SIW. The proposed method introduces two transmission zeros, effectively improving out-of-band suppression and enhancing the frequency selectivity required for the band-pass filter. Simultaneously, the design employs integrated passive device (IPD) technology to achieve miniaturization and high integration of the CSRR-SIW band-pass filter. The measurement results indicate that the designed band-pass filter achieves a minimum insertion loss of 3.18 dB within the pass band, with a shape factor of 2.1, a relative bandwidth of 18%, and dimensions of 2.74 mm×1.50 mm×0.54 mm. For a miniaturized band-pass filter using IPD technology, this design exhibits good shape factor.

China's reform and opening-up policies have brought profound changes in the social sphere.The living environment and socialization factors influencing contemporary youth have undergone dramatic transformations.As a result,young people's perception and evaluation of social issues or general phenomena have also changed significantly across generations,which in turn has considerable implications for their values and behavioral patterns.Based on longitudinal data from the 2013 and 2021 waves of the Chinese Social Survey(CSS),this study systematically analyzes the dynamic changes in Chinese youth's perceptions of social fairness,attention to social hot issues,social trust,and their evaluations over the past decade.The findings reveal that youth have become markedly more attentive and sensitive to these issues,though the implications of various indicators differ substantially.Their evaluations tend to be more rational and diversified,with increased inclusiveness,reflecting a trend toward value pluralism among today's youth.

Compositional Zero-Shot Learning (CZSL) defines categories as combinations of attributes and objects, requiring models to accurately recognize novel combinations not encountered during training. CZSL not only reduces dependence on labeled data but also explores the compositional generalization capabilities of deep learning models. Existing methods often neglect the appearance variability of attributes and objects in different compositional contexts, resulting in insufficient generalization. To address this, we propose a Context-Aware Decision Mechanism for CZSL. Leveraging the multimodal pretrained CLIP model as an encoder, our approach incorporates textual soft prompts, adapter fine-tuning, and feature denoising mechanisms to independently extract attributes and object features. Furthermore, we design a context-related gating network to adaptively fuse the predicted scores of attributes and objects during the decision-making phase. Experiments on four standard benchmark datasets demonstrate that our method significantly enhances CZSL performance, validating the effectiveness of context-aware decision-making in handling attribute-object combinations.

In view of the characteristics of short duration and subtle facial muscle movements of micro-expressions, a dual-stream Transformer-based micro-expression sequence recognition network (TSTMER) is proposed. Firstly, the TV-L1 optical flow method is used to calculate the optical flow information between the micro-expression sequence frames and the starting frame, in order to obtain the motion characteristics of facial muscles in micro-expressions. Secondly, a micro-expression feature extraction module is designed, which consists of two Transformer networks and a cross-attention mechanism. The two Transformer networks are used to extract spatial features of optical flow images and micro-expression images respectively, and the cross-attention mechanism fuses their spatial features into new ones. Then, a spatio-temporal attention mechanism is designed to enhance the weights of micro-expression local regions and sequence frames. Finally, Bi-LSTM is used to learn both the forward and the backward dependencies of micro-expression sequences in time dimension. Experiments on multiple micro-expression datasets demonstrate that the TSTMER network has significant advantages in UAR and UF1.

The transmission of wireless communication system is not able be guaranteed without an excellent filter. Microstrip filter is widely used because of its miniaturized volume and easy integration characteristics. However, high frequency will greatly increase the energy loss of microstrip filter, especially the radiation loss, which will increase the transmission loss and reduce the transmission efficiency. To solve this problem, a N78-band third-order bandpass microstrip filter is designed by combining multi-layer structure with defected ground structure (DGS), using three optimized I-shaped resonators and introducing transmission zero. According to the simulation, the radiation loss of the filter is less than 0.1 dB, the center frequency is 3.55 GHz, the fractional bandwidth is 22.3%, the return loss is -22 dB, and the area is 0.67λ_g×0.5λ_g. The filter can commendably reduce the radiation loss and realize the miniaturization, which has practical value.

For the purpose of solving the problem that particle swarm optimization has precocious convergence and will fall into local optimal solution with little hindrance, a particle swarm optimization algorithm with nonlinear inertial weight and t-distribution mutation is proposed. By improving the inertial weight and t-distribution mutation operators, the algorithm enhances the global search capability of the particles. Simulation results reveal that the improved algorithm has strong global optimization ability and effectively overcomes the premature convergence phenomenon.

To evaluate the effectiveness of science and technology talent policies, this paper selects 327 policy documents related to science and technology talent from the Yangtze River Delta and the northeast China regions over the past 12 years. Using Nvivo software, the study employs the “Projection-Post” comparative analysis method and entropy weight method to compare the policies in terms of policy tools, policy goals, and policy implementation, analyzing the mechanisms behind these policies. The findings show that the policy tool structure in northeast China is more imbalanced, while the Yangtze River Delta demonstrates a more complementary structure of policy tools. The implementation of science and technology talent policies in the Yangtze River Delta is more effective, whereas northeast China's implementation shows less sustainability and stability. Both regions perform best in demand-driven tools, though the internal contribution rates of policy tools differ. Over time, both regions have shifted from a substitutionary to a complementary use of supply-side and demand-side policy tools, reflecting an emphasis on the organic integration of policy tools. The policy goals of each region and phase are distinct, avoiding a “one-size-fits-all” approach. Both regions emphasize the cohesion between policy goals, forming a unified regional strategy, though mismatches between tool-goal combinations exist.

Technology licensing is a crucial means for firms to accelerate the diffusion of new technologies and realize their commercial value. To explore how firms implement technology licensing strategies in a competitive environment, this study constructs a duopoly competition model with two manufacturers, where one firm owns the innovative technology. By comparing the profits and outputs of both firms under three licensing modes—no licensing, fixed-fee licensing, and per-unit licensing—this study examines the technology owner's strategic choice of licensing and the diffusion of technology in a competitive market. The results indicate that when the attractiveness of the new technology market is low, technology licensing does not necessarily generate higher profits for the technology owner. While per-unit licensing can mitigate competition between firms, it is not always the optimal choice for the technology-owning firm. Additionally, the study finds that the greater the market attractiveness of the new technology, the more likely the technology owner is to adopt a fixed-fee licensing model.

This paper compares the family policies of East and West Germany, analyzing their effects on female employment rates and birth rates under different models. The study finds that West Germany's family policy tends toward the “re-familization” of childrearing, which has forced some professional women to delay or forgo childbirth. As a result, while female employment rates have increased, they have fallen into the trap of low fertility rates. In contrast, East Germany's family policy emphasized the “de-familization” of childrearing, which led to a simultaneous increase in both female employment rates and birth rates. Drawing on the experiences and lessons from both East and West Germany, China could build a “work-family friendly” family policy system to promote the long-term balanced development of its population structure.

The legal functions of the cross-border data“whitelist”mechanism include:expanding the“circle of trusted partners”for cross-border data flows,supporting data circulation under the Belt and Road Initiative,and facilitating equal,mutually beneficial,and progressive dialogue and negotiation between nations.In general,the mechanism follows an exploratory path characterized by a“from near to far”and“point-to-surface”approach.The“whitelist”mechanism should be positioned alongside other regulatory frameworks—such as security assessments,standard contractual clauses,and professional certifications—rather than serving as a substitute.The criteria for applying the“whitelist”should take into account several factors,including:the authority of public institutions in the recipient country to access personal data,the country's personal data protection policies and laws,mechanisms for implementing personal data protection,and relevant data protection treaties,agreements,or international commitments.

As a new economic paradigm following the agricultural and industrial economies,the digital economy has rapidly expanded,driving the growth of service sectors such as digital consulting,IT advisory,and demand analysis.This expansion has had a profound impact on China's economic structure and societal operations.Drawing on provincial panel data from 2014 to 2021,this study empirically investigates the relationship between digital economy development and the tertiarization of industrial structure,focusing on both output and efficiency dimensions.The results show that a one-standard-deviation increase in the digital economy index corresponds to average increases of 0.216 and 0.503 standard deviations in tertiarization output and efficiency levels,respectively.Mechanism analyses further reveal that digital economy development promotes industrial tertiarization indirectly through enhanced market integration and improved capital-technology efficiency.Heterogeneity analysis indicates that in central regions and highly urbanized areas,the digital economy significantly boosts both output and efficiency.In contrast,in western regions and less urbanized areas,the digital economy primarily drives output growth,while in eastern regions,it predominantly enhances efficiency

Based on panel data from 30 Chinese provinces between 2013 and 2022,this study employs the entropy method,kernel density estimation,Dagum Gini coefficient,spatial Markov chain,and standard deviation ellipse model to measure the development level and spatio-temporal evolution characteristics of new-quality productive forces in China.The findings are as follows:(1)Green and low-carbon development is a key indicator in assessing the level of new-quality productive forces.(2)Overall,the development level remains low and shows a decreasing spatial gradient from east to central,northeast,and then west China.(3)In terms of regional disparities,the overall inter-regional difference first narrowed and then widened.Intra-regional differences follow the pattern:west > east > central > northeast.(4)Regarding dynamic trends,the development of new-quality productive forces is on the rise nationwide and across all regions.Regional polarization has eased,although absolute disparities have widened.(5)From the perspective of spatial dynamics,the spatial distribution of development levels demonstrates a diminishing trend from the eastern coastal regions to the inland areas,with high-performing provinces primarily concentrated in southeastern coastal regions.

Aiming at the problemof excessive computational load in commonly used neural network classification models, this paper proposes a lightweight Convolutional Neural Network model that can be deployed in low-end embedded devices-a lightweight convolutional neural network model based on MobileNet and ShuffleNet. This model combines the characteristics of MobileNet and ShuffleNet, introduces Squeeze-and-Excitation channel attention mechanism and shortcut branch, and uses improved h-swish activation function to replace the original ReLU activation function, so as to improve both the recognition accuracy and the computational efficiency. Test results verify that the proposed lightweight convolutional neural network model based on MobileNet and ShuffleNet outperforms traditional lightweight convolutional neural network models such as MobileNet, ShuffleNet, EfficientNet, etc.

The detection of cattle targets in complex environments is a key issue in precise counting of cattle numbers based on machine vision. Due to factors such as occlusion caused by overcrowding of cattle and incomplete cattle individuals caused by their position at the edge of the camera, existing cattle target detection methods are not suitable for complex environments in breeding farms. This paper proposes a cattle detection algorithm in complex environments based on Swin-Transformer(SWT)-YOLOV5s network: first, we propose a backbone network SWT Backbone for cattle feature extraction, which is cascaded by double-layer shortcut(SC)-SWT and multi-layer convolution. Using the characteristics of SC-SWT module that focuses on global features, combined with the residual multi-layer convolution module that focuses on local features, we increase the depth and receptive field of the network, so that the model can fully extract both global and local features of cattle. Then, the feature fusion target detection head SWT Head, which is cascaded by not shortcut(NSC)-SWT and pyramid network, is proposed. Through the pyramid network, the feature pyramid of multi-scale fusion of global features and local features extracted from the SWT Backbone backbone network is constructed. Combined with the global receptive field of the NSC-SWT module Transformer and the CNN local receptive field of the C3 module, it enables the model to more accurately detect and filter the global features of cattle at high semantic levels and local features that represent the details of cattle in the feature pyramid, while efficiently removes feature interference from the background environment, and improves the detection accuracy of cattle targets in complex environments. Simulation experiments were conducted on the COWYCTC-1480 dataset collected in the laboratory. Compared with the widely used YOLOV5s and SSD algorithms, the accuracy, recall, and mAP of our method on the test set were 7.0%, 2.0%, and 11.1% points higher than the YOLOV5s algorithm,20.0%,8.0%,32.0%, and 29.3 points higher than the SSD algorithm, respectively.

The theoretical and experimental studies on the generation of asymmetrical transverse mode beams are carried out by employing an off-axis pumped Nd: YVO₄ laser with adjustable output reflectivities and an extra-cavity single-cylindrical lens. Numerical results show that the mode order of the output transverse modes is positively correlated with the off-axis displacement under the same intracavity loss. In addition, with the increasing of the cavity loss, the output modes become more asymmetrical. A corresponding experimental device is established, the experimental results show that as the output mirror reflectivities decrease from 95% to 81%, 60% and 50%, the transverse modes symmetries of 93%, 70%, 43% and 14%, are realized, respectively, which are consistent with the numerical simulation results. Furthermore, when the output mirror reflectivity is greater than 60%, the output power is positively correlated with the laser mode order number, otherwise the output power is negatively correlated with the mode order number. When the output mirror reflectivity is 60%, the output powers of the asymmetrical transverse modes exceed 1 W with a pump power of 6.36 W, and the maximum output mode can reach TEM_9,0. The experimental results are in good agreement with the numerical simulation results.

Aiming at the problems of current gait recognition methods in the recognition accuracy under cross-view conditions and performance under complex wearing conditions, this paper proposes a cross-view gait robot with global and local feature fusion (GLFF), which combines with the introduction of neural network hyperparameters to form the brain neural network for enhancing the network capabilities. Firstly, the power segmentation camera captures the movement information of the limbs and enhances the fine-grained characteristics of the network block space; and then, uses the set pooling layer (SP) to extract the color features with high viewing angle robustness, aggregate the gait at the time point of the asynchronous state, integrate a series of features and use the micro-action template builder to extract the micro-action features, and train to obtain GLFF. The CASIA-B data set of the Institute of Automation of the Chinese Academy of Sciences was used for experimental verification. The average recognition accuracy of this software under cross-view conditions reached 88.9%, and the recognition rate under simulated wearing conditions (CL) reached 79.5 %. This result can verify that fusion with local components can effectively reduce the influence of wearing factors on cross-view gait recognition.

Based on the SMIC 55 nm process, a 2.4 GHz phase-locked loop (PLL) with sub-sampling technology was designed to achieve low spurious and low jitter performance. In response to the issues of significant jitter and high in-band noise in traditional charge pump PLLs, the sub-sampling technique was introduced. Additionally, the three mechanisms affecting reference spurs in the sub-sampling PLL were investigated, and methods such as the Dummy sampler, isolation buffer, and a novel sub-sampling charge pump were proposed to reduce the level of reference spurs in the output signal. To avoid frequency locking errors and reduce lock-in time, a frequency locking loop incorporating a novel dead-zone circuit was employed. Considering the influence of the input signal on output noise, a variable duty cycle input signal buffer was designed to further decrease reference spurs. Simulation results demonstrate that under a 1.2 V power supply voltage, the sub-sampling PLL achieves a reference spur level of -87.72 dBc at a 50MHz frequency offset and an in-band phase noise of -126.3 dBc/Hz at 1 MHz. The overall power consumption of the system is 11.72 mW.

With the increasing size of financial assets , countries around the world are gradually becoming more transparent in their monetary policies, using central bank monetary policy communication as a new type of monetary policy instrument. To measure central bank communication more accurately, this paper proposes a dynamic topic model based branching processes, and systematically constructs a central bank communication index by combining the key wording method. At the same time, the paper uses the constructed communication index as a new type of monetary policy instrument to build a TVP-FAVAR model, which can extract potential macroeconomic information from a large number of macroeconomic variables, and its time-varying nature can also better reflect the dynamic regulatory effect of monetary policy. Experimental results demonstrate the effectiveness of central bank monetary policy communication,which can capture timely information on conventional monetary policy instruments and immediately respond to changes in interest rates and money supply, which can effectively smooth out output volatility and have a longer-term impact on the macroeconomy.

The traditional strawberry maturity prediction method mainly relies on the professional knowledge and experience of planting personnel to observe and detect the growth stages of different varieties of strawberries, and estimate the maturity of strawberries. However, this method is inefficient. In response to such issues, this paper is based on the national standard for strawberry maturity, and further refines the maturity of strawberries under this standard. We established a strawberry dataset, trained the YOLOv5 object detection model, and used this model to numerically analyze the maturity of strawberries. Numerical results were statistically analyzed, and the data was processed to obtain sample points. Finally, univariate polynomial regression, neural network regression, and DoubleBoltzmann fitting methods were used to regress the growth curve of red strawberry under specified growth conditions, and the results were compared and evaluated. Experimental results indicate that the proposed growth model for strawberries during their ripening period can predict and analyze the growth trend of strawberries throughout the entire process from fruiting to fruit ripening.

Aiming at the problem of limited energy stored in unmanned aerial vehicle (UAV), a resource allocation strategy for UAV-assisted edge computing in wireless powered communication networks is investigated. By deploying a laser beam director on the ground, sufficient energy can be provided for the UAV in a short period of time. Then, multiple ground terminals obtain energy from this UAV by radio frequency energy harvesting method and offload their computing tasks to the UAV with edge servers. The resource allocation problem is modeled as an optimization problem whose objective is to minimize the total energy consumption of the UAV subject to the constraints of energy and data causality, computational resources, and transmitting power. The suboptimal solution is obtained by introducing an imperialist competitive algorithm. Simulation results show that the imperialist competitive algorithm consumes less energy compared with the particle swarm optimization algorithm and the equal upload time allocation method.

Mined information from massive medical texts and extracting key named entities is crucial to improving the efficiency of medical information processing and dialogue understanding.Aiming at the difficulty of recognizing complex medical terms in named entity recognition tasks in the medical field, and the difficulty of comprehensive and accurate capture by a single feature method, a medical text named entity recognition method is proposed, which integrates multiple features of Pinyin, glyphs and lexicon information.This method employs pre-trained models to extract semantic vectors, constructs glyph vectors based on Chinese character strokes and radicals, along with tonal Pinyin vectors. It aligns medical dictionaries through bidirectional matching, dynamically fuses multi-modal features, and ultimately inputs them into a BiLSTM-CRF model to enhance entity recognition accuracy. The experimental results demonstrate that the proposed method significantly outperforms other comparative experiments in recognizing online doctor-patient conversation texts from the IMCS internet platform. The highest F1 score achieved on the test set reaches 92.5%, representing an improvement of up to 8.6 percentage points compared to the F1 scores of the other five comparative experiments (Base-BERT, MC-BERT, RoBERTa-wwm-ext, Medbert-Kd-Chinese, and Albert-Base-Chinese). Furthermore, ablation experiments also further verified the effectiveness of the fusion features in improving the recognition performance.