Abstract:To address the issue of an insufficient number of"bubbles"in the research on antimonotonicity in memristive chaotic systems,a four-dimensional memristive chaotic system with a multi-bubble cascade structure was proposed.A hyperbolic tangent memristor was introduced as a feedback element into a three-dimensional continuous autonomous system.The system was analyzed through numerical simulations and dynamical analysis tools such as bifurcation diagrams,and was also implemented on a field programmable gate array(FPGA).The results showed that as parameters varied,multiple periodic"bubbles"evolved into a Feigenbaum tree structure.Characteristics such as constant Lyapunov exponents,extreme multistability,and coexisting antimonotonicity were observed,demonstrating rich complex dynamical behaviors.Furthermore,the Matlab numerical simulation results were consistent with the FPGA hardware implementation results,confirming the physical realizability of the system.This study provided a new approach for investigating complex dynamical behaviors in memristive chaotic systems.

Abstract:Investigations were conducted on the gravitational dynamical properties of Kerr-like black holes surrounded by dark matter halos,with a focus on examining the influence of dark matter halos on the motion of spinning particles.The Mathisson-Papapetrou-Dixon(MPD)equation combined with the Tulczyjew spin supplementary condition(SCC)was employed to derive the effective potential of spinning particles,and the physical characteristics of their innermost stable circular orbit(ISCO)were analyzed.The results indicated that as the density and radius of the dark matter halo increased,both the orbital radius and angular momentum of the spinning particles' ISCO exhibited a monotonically increasing trend;in contrast,the energy showed significant non-monotonic variation characteristics:it first decreased with the increase of dark matter halo parameters,reached a minimum value,and then rose gradually.The research findings deepened the understanding of the gravitational interaction mechanism between black holes and dark matter halos,providing a theoretical basis for comprehending their complex dynamical processes.

Abstract:To address the issue of low classification accuracy in the traditional encrypted traffic-bidirectional encoder representations from Transformer(ET-BERT)model,structural optimization and performance improvements were conducted based on a reproduced version of the original ET-BERT model.Firstly,a learning rate warmup(Warmup)strategy was introduced to smooth the training process and enhance convergence stability.Secondly,a convolutional neural network-BERT(CNN-BERT)fusion module was designed to strengthen local feature extraction while retaining the global modeling capability of the Transformer.Finally,a dropout layer was added to reduce overfitting and improve model generalization.Experiments were performed on the lightweight encrypted traffic dataset.The results showed that the improved ET-BERT model achieved increases of 4.70 and 4.36 percentage points in F1-score(F1)and accuracy(ACC),respectively,compared to the original ET-BERT model,leading to high-precision traffic classification.It was demonstrated that the improved ET-BERT model effectively enhanced the classification accuracy,thereby providing a reliable technical pathway for the optimization of encrypted traffic classification models.

Abstract:To investigate the interventional effect of selenomethionine(SeMet)in nonalcoholic steatohepatitis(NASH),a mouse model of NASH was established using a choline-deficient,amino acid-defined(CDAA)diet,followed by SeMet intervention via gavage.The results showed that SeMet intervention significantly alleviated hepatic steatosis and inflammation,and improved serum liver function and lipid profiles.It also promoted the recovery of beneficial gut microbiota,such as Akkermansia muciniphila and the Dubosiella genus,while suppressing the expansion of conditional pathogens like Enterococcus.At the molecular level,SeMet markedly upregulated the messenger ribonucleic acid(mRNA)expression of key hepatic antioxidant selenoproteins and reversed the CDAA diet-induced downregulation of short-chain fatty acid receptors.These findings suggested that SeMet might protect against NASH by cooperatively regulating the thioredoxin reductase(TXNRD)and glutathione peroxidase 4(GPX4)antioxidant pathways along with the"gut microbiota-short-chain fatty acid receptor"signaling axis,thereby providing a theoretical basis for its development as a functional dietary component.

Abstract:To explore the utilization strategy and survival mechanism of woody plant resources in special habitats,and to clarify the niche and interspecific association characteristics of dominant species in Ganxishan Forest Park,Lichuan City,Hubei Province,the research was carried out through field quadrat survey,combined with niche breadth,niche overlap,variance ratio method,chi-squared test,Pearson correlation test and Spearman rank correlation test.The results showed that:The importance value of Pinus massoniana was the largest,followed by Pinus armandii,and the niche breadth of Eurya breviflora was the largest,followed by Pinus massoniana,which were the main constructive species of the community.The importance value and niche breadth ranking are quite different.The niche overlap of dominant species was common,and the overlap index of 41.05%species pairs was greater than 0.50.The dominant tree species were significantly positively correlated.In the chi-squared test,97.89%of the species pairs were in a state of no association or no significant association.Pearson correlation test and Spearman rank correlation test were consistent with the results,and the interspecific association was weak.Most of the species in Ganxishan Forest Park were not strongly connected,and the competition of each dominant species for environmental resources was weak,and the community was in an unstable succession stage.

Abstract:To address the issue that bearing fault diagnosis methods were often sensitive to outliers and struggle to fully capture the underlying geometric structure of the multimodal data,a bearing multimodal fault diagnosis method based on hyperspace geometric structure perception(HGSP)was proposed.Firstly,the original multimodal features were uniformly mapped into a hyperspace constructed using the Poincaré ball model,enabling the characterization of latent nonlinear structural relationships among data.Then,a structure-preserving strategy based on cosine similarity was designed to perceive intra-modal semantic consistency and enhance inter-modal feature fusion.Finally,the method took into account the geometric differences and directional similarities among multi-modal data.It enhanced the robustness against outliers and the ability to perceive the geometric structure of the data,significantly improving the separability of fault categories and the diagnostic accuracy.Experiments were conducted on the Paderborn University(PU)bearing dataset and the self-developed experimental platform roadheader dataset.Compared with the locality preserving canonical correlation analysis(LPCCA)method,the average recognition accuracy of the HGSP method on the PU dataset was improved by 2.00 percentage points,and by 1.83 percentage points on the roadheader dataset.The results indicated that the method can effectively enhance the discriminability of fault categories and demonstrate practical value in bearing fault diagnosis.

Abstract:To address issues such as low hit ratio,discontinuous evaluation,and unstable migration of high-frequency time series data from the electron cyclotron resonance heating(ECRH)system in cache,a remote dictionary server cache optimization method based on dynamic heat evaluation and three-level partition(DHE-TP-Redis)was proposed.The data value was quantified through a dynamic heat calculation model,and resources were intelligently allocated via a three-level data classification strategy.Basic heat retention was introduced to ensure the continuity of data heat,while stable operation of the cache under high concurrency was achieved through gradual elimination and module collaboration.The results demonstrated that this method exhibited outstanding performance under three access patterns and continuous migration scenarios.In the burst concentration mode,the hit ratio of DHE-TP-Redis method increased by 0.8 and 26.0 percentage points respectively,compared with the least recently used(LRU)method and least frequently used(LFU)method.In the continuous migration scenario,its migration frequency was reduced by 5.7 times/h compared with LRU method and 6.7 times/h compared with LFU method,and the fluctuation of service response time was reduced by 4.4 percentage points compared with LRU method and 5.1 percentage points compared with LFU method.This method provided a stable solution for high-frequency time series data caching.

Abstract:To address the problems of insufficient fusion of multi-source behaviors and weak targeted preference adaptation in knowledge recommendation for information security courses,a knowledge recommendation method,namely,bidirectional long short-term memory-multi-head attention-fusion of student multi-source behavior data(BiLSTM-MA-FSBD)was proposed.Firstly,an integrated feature system covering dynamic time series and static correlations was constructed where multi-source behavior data of students were integrated and core behavioral features were extracted.Secondly,a BiLSTM network was designed to encode the dependency relationships of behavior sequences,and a MA mechanism was utilized to adaptively assign weights to behaviors,thus achieving accurate inference of learning preferences.Finally,a three-level knowledge graph for information security was built to quantify the dependency relationships among knowledge points,and personalized recommendations were implemented by combining preference matching degrees.The results indicated that the recommendation precision of the BiLSTM-MA-FSBD method was improved by 26.2 percentage points in comparison with the collaborative filtering(CF)method.This method could effectively adapt to the teaching characteristics of information security courses and the personalized learning needs of students,and it provided a feasible technical solution for solving the problem of accurate knowledge recommendation.

Abstract:To address the issues of missing global semantic correlations between events and argument roles and insufficient utilization of document information in document-level multi-event extraction,a document-level multi-event extraction model based on argument correlation and graph neural network(DEEACG)was proposed.Firstly,entities were obtained using bidirectional encoder representations from Transformers(BERT)module,and an entity co-occurrence prediction task was introduced to enhance semantic associations among entities.Secondly,learnable event proxy nodes were incorporated to construct a heterogeneous graph consisting of entities,contextual information,and proxy nodes.Through the graph neural network with feature-wise linear modulation(GNN-FiLM)and multi-head self-attention mechanisms,global interactions and semantic fusion among multiple events were achieved.Thirdly,event type detection was performed using a multi-layer perceptron.Finally,a dual-projection space was constructed to model argument correlations.The Bron-Kerbosch algorithm was applied to extract maximal cliques from the graph as candidate argument combinations,and multi-head attention was utilized for argument role classification.The results demonstrated that DEEACG model was evaluated on the Chinese financial announcements(ChFinAnn)dataset,where its performance in multi-event extraction tasks showed significant improvement.Compared with the relation-enabled document-level event extraction(ReDEE)model,the mean value of F1-score was increased by 2.1 percentage points.This study confirmed that DEEACG model could effectively capture semantic associations among multiple events and was suitable for document-level multi-event extraction tasks.

Abstract:To address issues such as timbre leakage,loss of prosodic details,and insufficient naturalness in generated speech,a multi-source feature decoupling voice conversion model based on diffusion models(MFD-VC)was proposed.Firstly,speech was decomposed into distinct subspaces with different attributes and was processed independently in the model,through which high-fidelity voice conversion was achieved via multi-attribute collaborative control.Secondly,a style encoder(SE)was designed to extract speaker timbre features from reference speech.Concurrently,a content feature encoder(CFE)and a wave network-based encoder(WN)were utilized to process content and fundamental frequency information,respectively.Finally,a multi-scale fusion 1 dimension(MSF1D)module was integrated into the U network diffusion model(UNetDiff)architecture to enhance multi-scale feature expression in skip connections,thereby allowing the generation of more detailed Mel-spectrum representations.The results showed that the MFD-VC achieved an equal error rate of only 12.2%on the Librispeech for text-to-speech(LibriTTS)dataset,while a high average opinion score of 4.35 for similarity was obtained.Moreover,on the voice cloning toolkit(VCTK)dataset,an equal error rate of only 15.2%and a high average opinion score of 4.24 for similarity were achieved.The MFD-VC delivered superior performance in terms of sound quality,similarity,and content clarity.

Abstract:To address the challenges of fine-grained feature perception and cross-region dependency modeling in forged image detection,a detection model named decoupled frequency refinement network(Decoupled-FR Net)was proposed.The model was designed with a spatial-channel decoupled attention mechanism,which effectively avoided the feature coupling problem in traditional hybrid attention and enhanced the independence and discriminative power of feature representation.A feature refinement module was introduced to enhance the perception ability of subtle tampering through hierarchical feature calibration and fusion.Combining with a context-aware mechanism,long-distance dependencies across regions were captured,thereby improving the overall detection performance.The results showed that,the accuracy and average precision of the Decoupled-FR Net model on the forensic synthetics(ForenSynths)dataset were improved by 2.4 and 0.5 percentage points,respectively,compared with the inter-patch dependency network(IPD-Net)model,and on the generative adversarial network(GAN)generation detection(GANGen-Detection)dataset,average precision was improved by 0.1 percentage points,compared with the frequency domain network(FreqNet)model.The model provided a new solution for fine-grained forged image detection and was of important application value in the field of multimedia forensics.

Abstract:To address the problem of decreased transaction rates caused by inaccurate price prediction on online freight platforms,a local context learning model with tabular prior-data fitted network(TabPFN)based on Shingling retrieval(ShinglingPFN)was proposed,which integrated Shingling retrieval and the TabPFN.Firstly,the w-Shingling retrieval algorithm was employed by the model to match the most similar orders to the predicted order from historical order data,and locally associated contextual data was constructed.Secondly,a pre-trained TabPFN model instance was loaded and initialized.The filtered order data were input into the model,and TabPFN was enabled to learn the association patterns between freight features and freight rates based on such contextual information.Finally,the freight rate prediction results of the freight sample were output.The results showed that the mean absolute error(MAE)metric of the ShinglingPFN model was decreased by 30.98%compared with that of the random forest(RF)model.The interpretability of the model was further enhanced through global sensitivity analysis.The ShinglingPFN model could provide decision support for platform optimization of pricing strategies.

Abstract:To address the practical challenges in improving vocational college teachers' digital literacy,such as fragmented platforms,redundant resource construction,and insufficient training targeting,and to facilitate the digital transformation of education,a closed-loop hierarchical empowerment model of"diagnosis-adaptation-reconstruction-evolution"was constructed based on the hierarchical task network(HTN)theory and combined with the practice of the"Zhike"platform.Teachers'digital literacy was divided into three levels(novice,competent,and expert)in the model,and corresponding three-level task chains(basic,intermediate,and advanced)were designed.The large artificial intelligence(AI)models and multi-source resources were integrated to form a four-dimensional linkage mechanism of"technology-resources-scenarios-evaluation".A controlled experiment was conducted with 1256 vocational college teachers as subjects.The results showed that the digital literacy of teachers in the experimental group relying on the platform was improved by 22.1%,which was 10.1 percentage points higher than the improvement in the control group.Among all dimensions,the"digital teaching application"dimension achieved the most remarkable progress,and 85%of the teachers recognized that the tasks were highly consistent with teaching needs.This study provided an operable and replicable path for the large-scale and precise improvement of vocational college teachers' digital literacy,offering strong support for the development of teachers'capabilities in the digital transformation of education.

Abstract:To investigate the impacts of land use changes on ecosystem services under different scenarios,the urban agglomeration of the middle Yangtze River was taken as a case study.Land use patterns under four scenarios in 2030 were simulated using the patch-generating land use simulation(PLUS)model.Meanwhile,the integrated valuation of ecosystem services and tradeoffs(InVEST)model was employed to evaluate the spatiotemporal changes in four key ecosystem services,namely,current and future carbon storage,habitat quality,water yield,and soil conservation in the region.The results showed that ecosystem services exhibited significant spatial heterogeneity,and the spatial distributions of different ecosystem service functions remained largely unchanged over time.It was revealed by multi-scenario comparisons that only the ecological protection scenario could achieve improvements in carbon storage and habitat quality.In the remaining scenarios,water yield was generally increased,but this was accompanied by decreases in carbon storage and habitat quality,while soil conservation fluctuated slightly across all scenarios.Land use change was identified as the key factor influencing ecosystem services,and different changes in ecosystem services were exhibited under different scenario simulations.The study indicated that scientific land allocation was an important approach to balancing urban development and ecological protection.

Abstract:Using Rorippa indica(L.)Hiern as raw material,the extraction process of polysaccharide from Rorippa indica by ultrasonic-assisted ionic liquid and its antioxidant activity were studied by response surface methodology.Firstly,on the premise of single factor experiment,Box-Behnken experimental design and response surface method were used to optimize the four key factors,namely ultrasonic temperature,liquid-solid ratio,ultrasonic time and ionic liquid concentration,and analyze their effects on the extraction rate and antioxidant activity of Rorippa indica polysaccharide.The results demonstrated that the ideal processing parameters were ultrasonic temperature of 70℃,liquid-solid ratio of 26∶1,ultrasonic time of 30 min and ionic liquid concentration of 2.9%.Under these optimized conditions,three repeated experiments were carried out to verify the prediction accuracy of the model.The measured extraction yield of polysaccharide from Rorippa indica reached 76.44 mg/g,and the deviation was only 0.12%compared with the predicted value of the model,which proved the high reliability and practicability of the regression model.The polysaccharide from Rorippa indica showed remarkable scavenging effect on DPPH and OH free radical,and the half-inhibitory concentration(IC50)values were 61.91 mg/L and 134.98 mg/L,respectively,which indicated that the polysaccharide from Rorippa indica had good antioxidant activity under the optimized technological conditions.The research offered a theoretical foundation for the in-depth development and use of Rorippa indica,and also provided a potential possibility for its application in functional food and medicine.

Abstract:To address the deficiency of existing research that predominantly focused on macro-mechanisms or single-scenario analyses and lacked a systematic evaluation from the perspective of rural residents'perceptions,a satisfaction evaluation index system encompassing three dimensions,namely,policy support,development factors,and regulatory intensity,was constructed based on a research sample of 648 rural residents in Chuzhou City,Anhui Province.The importance-performance analysis(IPA)model was employed to compare the perceived importance and actual satisfaction of each indicator.The results indicated that the values of importance for all indicators range from 3.98 to 4.11,while the satisfaction scores range from 3.84 to 3.97,showing that the perceived importance generally exceeds actual satisfaction,with the larger gap observed in the regulatory intensity dimension.These findings provided practical empirical evidence for improving rural digital financial services and strengthening financial support for rural revitalization in central agricultural regions.

Abstract:To address the challenge of balancing budget constraints and location privacy in mobile crowdsensing(MCS),a budget-constrained worker recruitment algorithm based on local differential privacy(LBWR)was proposed.Firstly,an adaptive circular perturbation(ACP)mechanism was designed,where perturbation centers and radii were dynamically optimized and the exponential mechanism was applied to generate perturbed locations,thereby protecting location privacy while maintaining data utility.Secondly,a quality-aware recruitment algorithm(QARA)was introduced,which employed weighted Voronoi diagrams to capture regional characteristics and utilized ant colony optimization to select worker sets under budget limitations,enabling coverage-quality-driven recruitment.Finally,experiments on the Gowalla location-based social network dataset(Gowalla)and driving directions based on taxi trajectories(T-drive)in Beijing datasets showed that the LBWR's consistently outperformed comparison algorithms in coverage quality,root mean squared error,and minimum cost maximum workload.Specifically,LBWR achieved superior performance in coverage quality,ACP effectively reduced the root mean squared error,and QARA outperformed comparison algorithms in both coverage quality and minimum cost maximum workload,demonstrating the LBWR's balanced capability between privacy protection and efficient worker recruitment.The LBWR effectively preserved worker privacy while ensuring efficient coverage allocation,highlighting its robustness and scalability.

Abstract:To investigate the effects of hormetic effect and dynamic radiotherapy on tumor cell dynamics,a tumor-immune model with hormetic effect was proposed,and a dynamic regulatory factor was introduced for study.The existence and stability of the fixed points of the model were analyzed.The conditions for period-doubling bifurcation were derived using the center manifold theorem and verified through numerical simulations.The results indicated that the hormetic effect during radiotherapy significantly influenced the number of tumor cells,and that the implementation of dynamic radiotherapy eventually stabilized the tumor cell population near the threshold level.This study provided a theoretical reference for the design of radiotherapy strategies that consider both hormetic effect and dynamic radiotherapy.

Abstract:To prepare the preparation of low-cost supercapacitor electrode materials,N/O/S self-doped foamy hierarchical porous carbon(denoted as POLC-650-0.5)with high specific surface area(1001.4m2/g)and high heteroatom content(N:2.8%,O:12.7%,S:0.4%)was directly synthesized in one step using discarded Platycladus orientalis leaves as raw material.The naturally contained calcium oxalate in the leaves was used as an in situ hard template and endogenous activator,while a small amount of KHCO3 was introduced as an external green activator.The synthesis was conducted under mild conditions with a mass ratio of KHCO3 to Platycladus orientalis leaves of 0.5∶1.0 and an activation temperature of 650℃.Its electrochemical behavior in a 2 mol/L Na2 SO4 electrolyte system was tested.The results indicated that the symmetric supercapacitor assembled with Platycladus orientalis leaves-derived carbon POLC-650-0.5 delivered a specific capacitance of 133.0 F/g at a current density of 0.5 A/g,and retained 84.6%of its initial capacitance after 10000 charge-discharge cycles at a current density of 5 A/g,demonstrating excellent energy storage performance.This study provided a feasible route for the high-value utilization of waste plant resources and the cost-effective production of high-performance porous capacitive carbon for supercapacitors.

Abstract:To address the insufficient generalization to unseen attribute-object compositions in compositional zero-shot learning(CZSL),a pixel-level feature modulation and text-guided refinement for compositional zero-shot learning(PFMTR)model was proposed to boost the recognition of novel compositions.Firstly,a pixel-level feature modulation(PLFM)module was devised,which employed a dual attention mechanism operating at both pixel-level and patch-level to enable fine-grained reassembly and semantic enhancement of image features.Secondly,a text-guided refinement(TGR)module was proposed,where textual features were used as queries and visual features as keys/values.This module leveraged cross-modal attention to compute semantic relevance weights,thereby dynamically guiding visual features with linguistic semantics and achieving cross-modal alignment.The results showed that,compared with other state-of-the-art models,the PFMTR model achieved outstanding performance on the University of Texas Zappos(UT-Zappos)dataset under the open-world setting,attaining 35.7%in the area under curve(AUC)and 49.7%in the harmonic mean(HM).This study demonstrated that the recognition of unseen compositions could be effectively enhanced by integrating pixel-wise local modulation with cross-modal semantic guidance,offering a viable technical route for CZSL in complex scenarios.