Data Assimilation （DA） is the core of modern Numerical Weather Prediction （NWP） systems. However， traditional methods are constrained by bottlenecks such as non-linear and non-Gaussian assumptions as well as high computational cost. Deep Learning （DL）， with its powerful non-linear fitting and end-to-end optimization capabilities， offers a revolutionary paradigm to address these challenges. Therefore， from a computer science perspective， this paper reviewed systematically and analyzed a three-stage evolutionary trajectory for the application of DL in meteorological DA： component replacement， paradigm shift， and system reconfiguration. First， it elaborated on the component replacement stage， in which DL serves as surrogate models to replace observation operators and correct model errors， thereby addressing traditional bottlenecks. Second， it provided an in-depth analysis of the paradigm shift stage， where generative models， represented by Variational Auto Encoders （VAEs） and diffusion models， reconstruct the assimilation problem as a probabilistic generation task. Third， it discussed the system reconfiguration stage， in which Artificial Intelligence （AI） meteorological foundation models such as FengWu and FuXi were adopted， and the assimilation module was embedded deeply， so as to construct a closed-loop system of integrated assimilation-forecast. Finally， the paper explored the core challenges in this field， including physical consistency， data dependency， and operational deployment， and prospected future research directions， such as neuro-symbolic fusion and open benchmark construction.

Blind image super-resolution reconstruction focuses on recovering Low-Resolution （LR） images into clear High-Resolution （HR） images in real-world scenarios. Most existing degradation estimation networks cannot accurately estimate the degradation information of LR images， and the traditional Residual-in-Residual Dense Blocks （RRDBs） in super-resolution networks have poor generalization ability， which fail to adapt to complex degradation types. To address the above issues， a blind image Super-Resolution reconstruction method by Integrating Degradation estimation and Dynamic residual-in-residual dense blocks （IDD-SR） was proposed. First， an integrating degradation estimation network was proposed and used for accurate estimation of image degradation information. Then， a Dynamic Residual-in-Residual Dense Block （DRRDB） was proposed to improve the universality of the super-resolution reconstruction network. Finally， accurate prediction of complex degradation information and high-quality super-resolution image reconstruction were achieved through end-to-end joint optimization. Experimental results on the Set14， BSD100， and DRealSR datasets demonstrate that the Peak Signal-to-Noise Ratio （PSNR） values of 4× super-resolution of IDD-SR are improved by 0.90 dB， 0.71 dB， and 0.97 dB， respectively， compared with those of PDM-SRGAN. It can be seen that the proposed method has superior detail recovery capability and artifact suppression effect.

In view of the problems of vascular pleomorphism on transverse sections and sampling imbalance in the process of detection， an improved Libra Region-Convolutional Neural Network （R-CNN） cerebral arterial stenosis detection algorithm was proposed to detect internal carotid artery and vertebral artery stenosis in Computed Tomography Angiography （CTA） images. Firstly， ResNet50 was used as the backbone network in Libra R-CNN， Deformable Convolutional Network （DCN） was introduced into the 3， 4， 5 stages of backbone network， and the offsets were learnt to extract the morphological features of blood vessels on different transverse sections. Secondly， the feature maps extracted from the backbone network were input into Balanced Feature Pyramid （BFP） with the Non-local Neural Network （Non-local NN） introduced for deeper feature fusion. Finally， the fused feature maps were input to the cascade detector， and the final detection result was optimized by increasing the Intersection-over-Union （IoU） threshold. Experimental results show that compared with Libra R-CNN algorithm， the improved Libra R-CNN detection algorithm increases 4.3， 1.3， 6.9 and 4.0 percentage points respectively in AP， AP₅₀， AP₇₅ and AP_S， respectively_{on the cerebral artery CTA dataset； on the public CT dataset of colon polyps， the improved Libra R-CNN detection algorithm has the AP， AP50， AP75 and APS increased by 6.6， 3.6， 13.0 and 6.4 percentage points， respectively. By adding DCN， Non-local NN and cascade detector to the backbone network of Libra R-CNN algorithm， the features are further fused to learn the semantic information of cerebral artery structure and make the results of narrow area detection more accurate， and the improved algorithm has the ability of generalization in different detection tasks.}

Aiming at the problem that Backtracking Search optimization Algorithm （BSA） is easy to premature and has weak local development ability in the optimization of kernel function parameters and regularization parameters of Least Square Support Vector Machine （LSSVM）， an integrated prediction model named CABSA-LSSVM was proposed. Firstly， the Cauchy population generation strategy was used to improve the diversity of historical populations， so that the algorithm was not easy to fall into the local optimal solution. Then， the adaptive mutation factor strategy was used to balance the global exploration and local development abilities of the algorithm by adjusting the mutation scale coefficient. Finally， the improved Cauchy Adaptive Backtracking Search Algorithm （CABSA） was used to optimize the LSSVM to form a new integrated prediction model. Ten UCI datasets were selected for numerical experiments. The results show that the proposed model CABSA-LSSVM has the best regression prediction performance when the population size is 80. Compared with the LSSVMs optimized by the standard BSA， Particle Swarm Optimization （PSO） algorithm， Artificial Bee Colony （ABC） algorithm and Grey Wolf Optimization （GWO） algorithm， the proposed model has the coefficient of determination increased by 1.21%-15.28%， the prediction error reduced by 6.36%-29.00%， and the running time reduced by 5.88%-94.16%. In conclusion， the proposed model has high prediction accuracy and fast computation speed.

Small Open Reading Frames （sORFs） in long non-coding RNA （lncRNA） can encode short peptides with length no more than 100 amino acids. Aiming at the problem that the features of sORFs in lncRNA are not distinct and the data with high reliability are not enough in short peptide prediction research， a Deep Forest （DF） model based on representation learning was proposed. Firstly， the conventional lncRNA feature extraction method was used to encode the sORFs. Secondly， the AutoEncoder （AE） was used to perform representation learning to obtain highly efficient representation of the input data. Finally， a DF model was trained to predict the short peptides encoded by lncRNA. Experimental results show that the accuracy of this model can achieve 92.08% on Arabidopsis thaliana dataset， which is higher than those of the traditional machine learning models ， deep learning models and combined models， and this model has better stability. In addition， the prediction accuracy of this method can reach 78.16% and 74.92% on Glycine max and Zea mays datasets respectively， verifying the good generalization ability of the proposed model.

By analyzing the security of a certificateless signature scheme without bilinear pairing proposed by Wang Y, et al. (WANG Y, DU W. Security analysis and improvement of certificateless signature scheme without bilinear pairing. Journal of Computer Applications, 2013, 33(8): 2250-2252), the result that the scheme can not resist forgery attack was pointed out and an improved scheme was proposed. The improved scheme enhanced the relationship of parameters in signature algorithm to resist forgery attack. The results of security analysis show that the improved scheme is proved to be existentially unforgeable against adaptive chosen message and identity attacks in random oracle model. The improved scheme is more efficient than the existing schemes for avoiding bilinear pairings and inverse operation.

There are two inefficient steps in GEP-Cluster algorithm: one is screening and aggregation of clustering centers and the other is the calculation of distance between data objects and clustering centers. To solve the inefficiency, an auto-clustering algorithm based on Compute Unified Device Architecture (CUDA) and Gene Expression Programming (GEP), named as CGEP-Cluster, was proposed. Specifically, the screening, and aggregation of clustering center step was improved by Gene Read & Compute Machine (GRCM) method, and CUDA was used to parallel the calculation of distance between data objects and clustering centers. The experimental results show that compared with GEP-Cluster algorithm, CGEP-Cluster algorithm can speed up by almost eight times when the scale of data objects is large. CGEP-Cluster can be used to implement automatic clustering with the clustering number unknown and large data object scale.

The existing parallel cost models are mostly devised for shared memory or distributed memory architecture, thus not suitable for heterogeneous multi-core processors. In order to solve the problem, a new parallel cost model for heterogeneous multi-cores was proposed. It described the impact of computing capacity, memory access delay and data transfer cost on parallel execution time of loops quantitatively, thus improving the veracity of accelerated parallel loop recognition. The experimental results show that the proposed model can effectively recognize the accelerated parallel loops. Using its recognition results to generate parallel codes can improve the performance of parallel programs on heterogeneous multi-core processors significantly.

Dynamic trusted measurement is a hot and difficult research topic in trusted computing. Concerning the measurement difficulty invoked by the dynamic nature of operating system kernel, a Dynamic Trusted Kernel Measurement (DTKM) model was proposed. Dynamic Measurement Variable (DMV) was presented to describe and construct dynamic data objects and their relations, and the method of semantic constraint was proposed to measure the dynamic integrity of kernel components. In DTKM, the collection of memory data was implemented in real-time, and the dynamic integrity was verified by checking whether the constructed DMV was consistent with semantic constraints which were defined based on the security semantics. The nature analysis and application examples show that DTKM can effectively implement dynamic measurement of the kernel and detect the illegal modification of the kernel dynamic data.