Aiming at the problem that existing methods could not fully capture the intrinsic structure of data without considering the higher-order neighborhood information of the data, we proposed an unsupervised feature selection algorithm based on graph filtering and self-representation. Firstly, a higher-order graph filter was applied to the data to obtain its smooth representation, and a regularizer was designed to combine the higher-order graph information for the self-representation matrix learning to capture the intrinsic structure of the data. Secondly, l_2,1 norm was used to reconstruct the error term and feature selection matrix to enhance the 
robustness and row sparsity of the model to select the discriminant features. Finally, an iterative algorithm was applied to effectively solve the proposed objective function and simulation experiments were carried out to verify the effectiveness of the proposed algorithm.

Aiming at the problems existing in few-shot learning at present, we designed a new network structure and its training method to improve the few-shot learning. The  convolution network and multi-scale slide pooling method were used to enhance feature extraction in the feature embedding part of the network. The main structure  of the networks was the Siamese network  to facilitate learning semantics from small sample data through comparison between samples. The training method  of the framework adopted nested level parameter updating to ensure the stability of convergence. Compared with the common visual model and 
few-shot learning methods, the experimental results  on two classical few-shot learning datasets show that the method significantly improves the  accuracy of  few-shot learning, and  can be used as a solution  under the condition of insufficient sample.

Aiming at the shortcomings of existing methods in the image segmentation performance of small and medium-sized organs in the abdomen, we proposed  a network model based on local and global parallel coding  for multi-organ image segmentation in the abdomen. Firstly, a local coding branch was designed to extract multi-scale feature information. Secondly, the global feature coding branch adopted the  block Transformer, which not only captured the global long distance dependency information but also reduced the computation amount through the combination of intra-block Transformer and inter-block Transformer. Thirdly, a feature fusion module was designed to fuse the context information from two coding branches. Finally, the decoding module was designed to realize the interaction between global information and local context information, so as to better compensate for the information 
loss in the decoding stage. Experiments were conducted on the Synapse multi-organ CT dataset, compared with the current nine advanced methods, the average Dice similarity  coefficient  (DSC) and Hausdorff distance (HD) indicators achieve the best performance, with 83.10% and 17.80 mm, respectively.

The hydrogel was prepared by free radical polymerization and freeze-thaw cycle using hydroxypropyl acrylate， ι-carrageenan，  polyvinyl alcohol and calcium chloride as raw materials. We tested and analyzed the mechanical property，  self-recovery property，  swelling property，  self-healing property，  anti-fatigue property and the conductivity property of the hydrogel. The results show that the hydrogel has high mechanical strength, the fracture stress is 625.77 kPa， the fracture elongation is 604.48%, the swelling rate of hydrogel is low, and has excellent self-recovery property，  anti-fatigue property  and  conductivity property.

Aiming at the problems of the limitation of obtaining global features in image recognition and the difficulty in improving recognition accuracy, we proposed  an image recognition method based on the lightweight model of random augmentation Swin-Tiny Transformer.  The method combined the random data augmentation based enhancement (RDABE) algorithm to enhance image features in the preprocessing stage, and adopted the Transformer’s self-attention mechanism to obtain more comprehensive 
high-level visual semantic information. By optimizing the Swin-Tiny Transformer model and fine-tuning the parameters on a maize disease dataset, the applicability of the algorithm was verified on maize diseases in the agricultural field, and more accurate disease detection was achieved. The experimental results show that the lightweight Swin-Tiny+RDABE model based on stochastic 
enhancement has an accuracy of 93.586 7% for maize disease image recognition. The experimental results compared with the excellent performance lightweight Transformer and convolutional neural network (CNN) series models with consistent parameter weights show that  the accuracy of the improved model is higher than that of the  Swin-Tiny Transformer, Deit3_Small, Vit Small, 
Mobilenet_V3_Small, ShufflenetV2 and Efficientnet_B1_Pruned models by 1.187 7% to 4.988 1%, and can converge rapidly.

Aiming at the problems  of low target segmentation rate and feeble correlation of image context feature information under multi-scale categories in image semantic segmentation, we proposed an image semantic segmentation model that fused attention mechanism and multi-scale features. The model used the improved  atrous spatial pyramid pooling to increase the segmentation of multi-scale targets, used the attention refinement module to capture context information to guide feature learning, and added feature fusion  based on attention mechanism to  supervise the learning of important channel features, guide the fusion of high-order and low-order features, so as to improve  the generalization capability of the model. The simulation results on the Cityscapes dataset show that the mean intersection over union of the model is 1.14% higher than that of DeepLab v3+, which proves that the model has good robustness.

We established a method based on liquid  chromatography-high resolution tandem mass spectrometry (LC-HRMS/MS) to detect A1 and A2 β-casein in milk. Caseins were extracted from milk by isoelectric point precipitation method,  after trypsin hydrolysis,  the hydrolysates were analyzed by LC-HRMS/MS method. The peptide containing varied amino acid moiety at the 67th position was selected as characteristic species to  distinguish A1 and A2 β-casein. Casein was fully dissolved in Tris-HCl buffer solution. The  acetonitrile was added to the enzymatic hydrolysis system to improve the enzymtic hydrolysis efficiency.  High resolution mass spectrometry and tandem mass spectrometry were used to eliminate the interference signal of impurity. Four kinds of common milk  and two kinds of  A2 milk  were tested by this method. The results show that the content of A1 β-casein is about 2—4 times that of  A2 β-casein in common milk,  while a small amount of  A2 milk samples contain trace A1 β-casein. This method can detect A1 and A2 β-casein in milk with advantages of high reliability and sensitivity,  which provides support for A2 milk source screening and quality control.

Based on the advantages of   high sensitivity,  simple operation and time-saving in   the detection of metal ions by fluorescent analysis.  We designed and synthesized a porous metal-organic framework  crystal fluorescent probe with amino modification. The experimental results show that the   compound is a two-fold interpenetrating  structure  of pcu framework,  and the amino groups on the organic ligands are freely distributed in the channels. The compound has excellent thermal  stability and water stability, and can be used as fluorescence enhanced probe to detect Ag⁺ in aqueous solutions,  with the detection limit of 0.76 μmol/L.

We consider a variant of the classic problem of  the traveling salesman problem (TSP) in combinatorial optimization problem: 
 the close enough traveling salesman problem (CETSP).  Firstly, we comprehensively introduce the history, solving methods, and algorithms for both TSP and CETSP, including exact algorithms (such as branch and bound method, linear programming) and heuristic algorithms (such as particle swarm optimization, greedy algorithms, etc.). The TSP requires finding the shortest path to visit each city  once and return to the starting point given a list of cities and distances. CETSP is a generalization of TSP, allowing the visiting point for each target to be chosen from within a specified neighborhood, rather than  exact location. It is  suitable for practical applications that can  tolerate errors, such as logistics distribution, intelligent transportation, and wireless sensor networks, etc. CETSP has higher flexibility and adaptability, which can significantly reduce computational resources and time consumption, particularly for large-scale problems with greater advantages. Secondly, we introduce  the potential  of CETSP in practical applications, especially in logistics, industrial manufacturing, traffic planning, information and communication, offering effective solutions for improving efficiency, reducing costs, and promoting intelligent decision-making. Finally, we have identified some future research directions for CETSP.

Mesenchymal stem cell exosomes (MSC-Exos) are a class of nanoscale vesicles with great potential in experimental research  and clinical applications. They contain a variety of biomolecules,  including miRNA,  mRNA,  proteins and lipids,  and have the function of  mediating  cell signaling and participating in regulation of receptor cells. Based on the important role of  MSC-Exos, we review the significant effects of  MSC-Exos in promoting tissue repair,  immune regulation and neuroprotection from the research progress and clinical applications,   especially in the treatment of autoimmune diseases,  neurodegenerative diseases,  cardiovascular diseases and tumors. We analyze a series of unsolved problems and application popularization challenges in its clinical application,  including elucidation of mechanism of action,  separation,  extraction and purification technology,  formulation of standardized production rules,  determination of dosage and administration route,  enhancement of stability,  and reduction of immunogenicity. This provides a basis for addressing these limitations to achieve widespread clinical application of MSC-Exos.

Aiming at the problem of network anomaly cause analysis in communication networks, we proposed  a method of alarm data feature analysis based on decision trees.  Firstly, the alarm data generated by network devices were mostly untagged data, and the alarm data was analyzed based on the correlation analysis of geographical and temporal features to obtain root  information and add tags, and  then the  alarm data  features were multi-dimensional. Secondly, the importance of various features and their impact on accuracy were analyzed by the decision tree method, and  the computational burden on time was reduced by pruning.  The experimental results of alarm data in each region show that the alarm compression rate is about 70% and 90% after pre-processing and correlation analysis, respectively. The impact of alarm object type and alarm logic classification features on network anomalies is stable, and  the importance degree is above 0.25, among them,  the root alarm mostly belongs to communication alarm in the alarm logic classification. This method can help network managers to identify the main alarm data features to a certain extent, and  recover the network first according to  the data features.

Ag-CDs nanoparticles with core-shell structure were obtained by reducing Ag ion with carbon dots (CDs). The results show that the diameter of Ag core is about 30 nm, and the thickness of surface carbon film is about 2.2 nm. Ag-CDs nanoparticles have excellent peroxidase-like activity and surface enhanced Raman scattering (SERS) activity, which can catalyze the oxidation of 3,3′,5,5′-teramethylbenzidine (TMB). Because uric acid can reduce the oxidized TMB, the concentration of uric acid can be detected by colorimetry. Taking Ag-CDs nanoparticles as the SERS substrate, and the concentration of uric acid can be detected by SERS spectrum. The collaborative detection of the two methods can greatly improve the reliability of the detection.

Aiming at the time-consuming and labor-intensive problem of obtaining large-scale multi-hop question and answer training dataset, we  proposed a multi-hop question generation model based on the contrastive learning idea. The model was divided into the generation phase and the contrastive learning scoring phase. In the generation phase, candidate multi-hop questions were generated by executing the inference graph. In the contrastive  learning scoring phase, candidate questions were scored and sorted through a candidate question scoring model without reference question based on the contrastive learning idea, and the best candidate question was selected. This model had to some extent narrowed the gap between unsupervised methods and manual annotation methods, effectively alleviating the problem of lacking a multi-hop question and answer dataset. The experimental results on HotpotQA dataset show that the multi-hop question generation model based on contrastive learning can effectively expand the training data and greatly reduce the cost of manually labeling data.

Aiming at the problem of large amount of work on the value of clustering number K in partition clustering algorithm, we proposed a new Canopy+ algorithm. The proposed algorithm can predict the clustering number K and improve the clustering efficiency on the premise of ensuring the accuracy.

This article implements the algorithm of using a random eventthe mouse movement tracks to produce real random numbers. Compared with the usual randomnumberproducing algorithm based on mouse, this one enjoys a higher randomicity in the process of sampling and getting original data. In addition, it does not need the extra circuit or facilities that other physical processes, the algorithms producing real random number. As a result, its costs will be reduced and the problem of a high expanse to produce real random numbers will be solved. The result of testing the uniformity and independence of the random numbers generated shows that the numbers produced by this algorithm have a fine statistical property. Moreover, the result of testing its program execution time proves that its time expending is very short.

Aiming at  the non-selective expansion  and training deficiencies of the DoubleMix algorithm during data augmentation, we proposed a supervised contrastive learning text classification model based on double-layer data augmentation, which effectively improved the accuracy of text classification when training data was scarce. Firstly, keyword-based data augmentation was applied to the original data at the input layer, while selectively enhancing the data without considering sentence structure. Secondly, we  interpolated  the original and augmented data in the BERT hidden layers, and  then send them to the TextCNN for further feature extraction. Finally, the model was trained by using Wasserstein distance and double contrastive loss to enhance text classification accuracy. The comparative experimental results on SST-2, CR, TREC, and PC datasets show that the classification accuracy of the proposed method is 93.41%, 93.55%, 97.61%, and 95.27% respectively, which is superior to classical algorithms.

Aiming at the problem that most traditional methods for multidimensional time series analysis relied on manually establishing temporal dependencies to explore the  implicit rules  in historical data, we proposed  an autoregressive neural network method. Firstly, the neural network composed of convolution neural network (CNN) and bidirectional long short-term memory (LSTM) was used to capture the complex dependencies existing in multidimensional input features and time series, and the linear relationship was extracted by combining the traditional autoregressive method. Secondly,  compared with several classical models on two datasets in different domains, the experimental results showed that the model had the best prediction performance and could  successfully capture the repeated patterns in the data. Finally, the  ablation experiments verified the efficiency and stability of the model framework.

In order to overcome the shortcomings of current image segmentation algorithms, such as large segmentation error and long segmentation time and inability to segment online image, we proposed an image segmentation algorithm based on combination of clustering analysis and active contour model. Firstly, the original image was denoised and roughly segmented by clustering analysis algorithm. The roughly segmented result was taken as the initial contour line of the active contour model. Secondly, the active contour model was used to fit the contour of different regions of the image according to the initial contour line to realize the fine segmentation of the image. Finally, it was compared with clustering analysis algorithm, active contour model and the current classical image segmentation algorithm. The experimental results show that the proposed algorithm overcomes the shortcomings of the current image segmentation algorithm, improves the efficiency and accuracy of image segmentation, it is insensitive to noise and has strong robustness. The overall image segmentation effect is significantly better than that of the contrast algorithm.

On the basis of analyzing  the commonalities between the correlation of Traditional Chinese Medicines and the relationship of data nodes of the graph structure, we  transfered the links of the drugs in a prescription into the graph structure data according to the rules. Dealing  the structured prescription data with CloseGraph, an efficient frequent closed subgraph mining algorithm, we  got frequent closed graph with a specific function of the graph structure. And then we obtained the core drug combinations and the forms of the combinations with a decisive effect on a specific disease for Traditional Chinese Medicines, providing a valuable scientific basis on figuring out the principles among diseasesyndromeformula. We successfully introduced the graph mining strategy into the field of Chinese Medicine research, providing a new idea and a more solid scientific theoretical foundation for both the prescription study and the future development of Traditional Chinese Medicines.

Aiming at the problem of  static resource allocation of the Storm platform, we proposed a distributed adaptive elastic resource allocation strategy,  which could  optimally meet the resource requirements of applications. Based on this strategy, combined with the resource allocation mechanism, application programming interface and user interface parameters of Storm, an elastic resource allocation component deployed in Storm was implemented to support adaptive and dynamic adjustment of application resources. The experimental results show that on the real stream data set, compared with the middle-value dynamic resource allocation strategy and the static resource allocation strategy of Storm, this distributed optimal strategy has advantages in throughput, loss rate and resource utilization. Meanwhile, this adaptive elastic resource allocation component can well interact with the Storm system, providing a reference solution for the development of other elastic resource scheduling components.

Firstly, based on the combination of sentence organization information and the influence of rules, partofspeech and word order on the syntax analysis system, we proposed a rulebased sentence analysis and recognition algorithm, which could quickly identify correct sentence patterns in large amounts of text. Secondly, based on the sentence analysis and recognition algorithm, we proposed a hierarchical sentence analysis and recognition algorithm based on the combination of rules and syntax to improve the accuracy of the hierarchical sentence recognition error detection. The experimental results show that the average accuracy rate and average recall rate are 8465% and 7715%, respectively, compared with the rulebased sentence recognition algorithm respectively improve 11.79% and 14.48%. It proves the feasibility of the analysis and recognition of hierarchical sentence combining rule and syntax.

We considered a class of sum of squares (SOS) convex polynomial optimization problems with spectrahedral uncertainty data in both objective and constraint functions. Firstly, an alternative theorem for SOS-convex polynomial system with uncertain data was established in terms of SOS conditions. Secondly, we introduced a SOS relaxation dual problem for this SOS polynomial optimization problem and characterized the robust weak and strong duality properties between them. Finally, a numerical example was used to demonstrate that the SOS relaxation dual problem could be reformulated as a semidefinite  programming problem.

In order to study the key genes and key metabolic pathways after the sgf73 gene was knocked out in Schizosaccharomyces pombe,  the wild\|type yeast strains and sgf73Δ  gene-deficient strains were sequenced and bioinformatics analyzed by RNA-Seq sequencing technology, and the GO and KEGG functional enrichment analysis were carried out.  The results show that in the sgf73Δ gene-deficient strains,  there are 1 834 highly expressed genes,  including 6 extremely highly expressed genes,  and 1 714 differentially expressed genes,  of which 934 genes are up-regulated and 780 genes are down-regulated. The  sgf73 gene knockout leads to  abnormal changes in cellular metabolism and transmembrane transport. The  down-regulation of cki1,cki2 and cdc25 genes involved in cycle regulation in the MAPK signaling pathway leads to prolongation of mitotic time in sgf73Δ strain,  the down-regulation of regulatory cytoskeleton rgf2,rho1 and stt4 genes leads to abnormal contraction  of actin ring of  sgf73Δ strain.

Aiming at the problems of the complicated underwater image imaging environment resulted in the subsequent image analysis often being affected by color bias and other factors, we proposed a deep convolutional neural network image restoration method based on multi-scale features and triple attention multimodal fusion. Firstly, the deep convolutional neural network introduced the image multi-scale transformation feature on the basis of extracting the image spatial feature. Secondly, by using channel attention, supervised attention and non-local attention, the scale correlation and feature correlation of image features were mined. Finally, by designing a multimodal feature fusion mechanism, the above two types of features could be effectively fused. The proposed method was tested on the open underwater image test set and compared with the current mainstream methods. The results show that this method is superior to the comparison method in quantitative comparison such as peak signal-to-noise ratio and structural similarity, as well as qualitative comparison such as color and details.

Aiming at the problem that the traditional neural network proportionalintegraldifferential (PID) controller and linear quadratic regulator (LQR) optimized PID controller had long recovery time and poor antiinterference for speed control of brushless direct current motor， we proposed a BP neural network PID controller optimized by LQR for speed control of brushless direct current motor. Firstly, BP neural network was used to adjust the PID gain to improve the dynamic adaptability and robustness of the controller. Secondly, LQR was used to optimize the optimal output of BP neural network, which was closer to the target PID gain. The simulation results show that the controller can effectively improve the response speed, reduce the steadystate error and enhance the antiinterference ability.

We discussed a class of  inverse problem of  time fractional order diffusion equations by using the fractional order Tikhonov regularization method. We gave selection criteria for the regularization parameter under  prior and posterior conditions, and gave rigorous  proofs of the convergence of the method. Numerical experimental results show that  the fractional order Tikhonov regularization method is effective in solving this problem.

We combined the traditional two memory-based collaborative-filtering methods and proposed a data-based personalized mixed recommendation method for GitHub projects. The method could not only calculate the similar users dynamically to ensure the personalized recommendation, but also obtain the recommendation quality comparable to the item-based method with only small scale of similar users. At the same time, the method solved the data sparsity and cold boot problems of the original method in the face of GitHub, a data set of users and projects of an order of magnitude but with low degree of crossover to some extent by establishing inverse table and using K-means classification. By comparing with the
traditional method, we verified the effectiveness and superiority of the proposed method.

Aiming at the problem that the traditional matrix-completion unconstrained optimization model had poor robustness in dealing with missing matrices damaged by singular noise, we proposed an adaptive robust matrix completion method. In this method, truncated kernel norm was used as the low-rank approximation of the rank function in the objective function, and the F-norm robust to singular noise was used as the loss term to recover the missing values in the matrix, so as to reduce the influence of outliers on the algorithm and improve the recovery accuracy. In the process of solving this model, a dynamic weight parameter was introduced by using convex optimization technique, which could be used to adjust the next update adaptively according to the current recovery error when updating the recovery value, and then an effective iteration method was established to solve the optimization problem. The experimental results show that the algorithm has better robustness and accuracy when dealing with matrices damaged by singular noise, so that better image restoration effects can be obtained.

Graphitic carbon nitride (g-C₃N₄) sheets were synthesized via the pyrolysis of melamine at 1 000 K under vacuum. X-ray diffraction patterns strongly indicate that the synthesized carbon nitride was g-C₃N₄. Transmission electron microscopic images indicate that the product was mainly composed of graphitic carbon nitride sheets. The chemical bonding of the sample was investigated by X-ray photoelectron spectroscopy and Fourier transform infrared spectroscopy. 

Hybrid particlefield molecular dynamics simulation method was used to simulate the selfassembly of block copolymer in bulk with large system (about 56 nm) and long time (100 μs) in order to avoid the finite size effect of the small system. The selfassembled structure of  diblock copolymer and triblock copolymer in bulk has been obtained by adjusting the number of blocks, the block ratio, and the interaction parameters between different blocks. The results show that the simulation results are consistent with the results of the selfconsistent field theory (SCFT), the Monte Carlo method (MC), and the dissipative particle dynamics method (DPD).