Abstract：Sparse Matrix-Vector Multiplication (SpMV) is an important linear algebraic subroutine in Matrix numerical computation. Vectorized Compressed Sparse Row (VCSR) sparse matrix compression format is proposed by studying the load balancing and memory access frequency of SpMV algorithm. This format adjusts the data load of each thread according to the statistical characteristics of the distribution of each line of non-zero elements to prevent the problem of thread divergence, and improves the computational performance of SpMV flow based on the strategy of fast segmented summation and the vectorization method. By using the Sparse matrix of the University of Florida as the test set, the performance of the GPU is tested, and the average performance improvement is 10% to 30%, and the maximum performance is 50% compared to the CSR5 (Compressed Sparse Row 5) format.

Abstract：Quantum computing has shown great potential in addressing traditional computational challenges, but due to its high error rates and noise issues, classical simulation has become an essential tool for verifying its performance. However, the superposition and entanglement properties of quantum systems pose significant challenges for simulation, especially when memory is limited. Although circuit cutting methods can decompose large-scale quantum circuits into smaller computational tasks to reduce computational load, previous research primarily focused on their application to quantum computers, without fully considering their effectiveness in quantum circuit simulation. This study fills that gap by proposing an optimization scheme based on a heuristic cutting algorithm and subcircuit state vector reuse to address memory limitations in simulations. By incorporating global computational cost considerations and an integer programming model, the heuristic method proposed in this paper not only optimizes the cutting process but also combines subcircuit state vector reuse to reduce redundant calculations and memory usage. Experimental results show that compared to current popular circuit cutting methods, the proposed approach significantly improves simulation speed while reducing memory requirements, effectively addressing the challenges in quantum circuit simulation. The overall average speedup achieved 46%.

Abstract：This paper proposes a gene heuristic multi sequence alignment algorithm based on the Yarn cloud platform. Establish a nucleic acid replacement equivalence matrix as a genetic heuristic mathematical model, construct the Yarn cloud platform logical architecture, and classify and save the data through steps such as gene data preprocessing, gene data storage, gene data alignment, gene data management, and gene data analysis. Divide long sequences with high error rates, and obtain multiple shorter gene fragments. Implementing localization on different fragments, generating variable length seeds, constructing skeletons and filling gaps, can achieve gene heuristic multi sequence alignment. The results show that the designed algorithm reduces processing time on different datasets, and the sum of pairs (SP) score for multi sequence alignment is higher. This experiment verifies the practicality of the multi sequence alignment method.

Abstract：With the explosive development of big data, sparse matrix has become an important part of machine learning and edge computing. In the field of machine learning, sparse matrix of data sets can not only save information but also save memory, which has become an inevitable trend. Sparse matrix vector multiplication (SpMV) is the core of sparse matrix computation. The space complexity and time complexity of its iterative solution process have important research significance. Analyze the compression format of sparse matrix C00, CSR, ELLPACK and DIA, change the sparsity of sparse matrix and the distribution of non-zero elements, and conclude that the SpMV read by COO and calculated by CSR is more universal. Utilizing the VLIW instruction architecture of C66x, using software pipelining to manage SpMV_CSR algorithm for instruction parallel optimization, utilizing SIMD single instruction multiple data instruction set for SpMV_CSR algorithm completes data parallel optimization. The experimental results indicate that the optimized SpMV_CSR algorithm has an average acceleration ratio of over 5 times compared to before optimization.

Abstract：Current artificial visual systems still struggle to handle real-world scenarios involving high-speed motion and high dynamic range scenes. Event cameras have the capability to address these challenges due to their low latency and high dynamic range for capturing fast-moving objects. However, reconstructing events into videos while maintaining their speed presents a challenge due to the highly sparse and dynamic nature of event data. Therefore, this paper proposes an event stream reconstruction algorithm based on Transformer residual networks and optical flow estimation. By jointly training optical flow estimation and event reconstruction, a self-supervised reconstruction process has been achieved. Additionally, deblurring preprocessing and subpixel upsampling modules are introduced to enhance the quality of reconstruction. Experimental results demonstrate that the proposed approach effectively improves the reconstruction quality of event streams on public datasets.

Abstract：Currently, with the large-scale integration of renewable energy into the distribution network, the intermittency and randomness of renewable energy sources such as solar and wind power inevitably cause fluctuations in the distribution network. Considering the characteristics of renewable energy generation power and electricity load in the power grid over time, a load prediction and optimization method based on wavelet transform and neural network for renewable energy access to the distribution network is proposed. Firstly, the grid operation data are collected, and the wavelet transform is used to process the collected data to obtain the feature parameters of local scale and frequency decomposition. A neural network is established. Then the feature parameters obtained after the wavelet transform are trained to obtain a model capable of predicting the load, according to which the power generation of renewable energy sources can be adjusted in time to maintain the dynamic balance between the supply and demand sides of the distribution network. The results show that the proposed method can effectively predict the load and regulate the power generation by observing the load in advance to ensure the stability of power consumption in the distribution network and simultaneously maximize the use of renewable energy.

Abstract：In recent years, face recognition has been a hot topic in society. Compared to contact-based recognition methods such as fingerprint recognition, face recognition offers the advantage of being contactless. In the field of deep learning, traditional convolutional neural networks do not achieve high enough accuracy or speed for face recognition. Therefore, this paper proposes a face recognition algorithm using the AlexNet convolutional neural network. Experimental results show that AlexNet provides higher accuracy and more stability in face recognition compared to traditional convolutional neural networks.

Abstract：To improve the security performance of SoC systems, this paper presents a design of a secure bus bridge circuit based on the ARM architecture. The circuit consists of a bridge module and an access control function module. Simulation results indicate that the circuit is capable of performing protocol conversion from AHB to APB under arbitrary multiples of AHB and APB clock frequencies. It also supports flexible configuration of access permissions and protection modes for master and slave devices. This prevents unauthorized access and malicious tampering of peripheral register data, thereby improving the security performance and applicability of the SoC system.

Abstract：The fault of higher level of IF-Noise was successfully handled for a receiver. The result shows that individual differences exist within the RF module cavity. It causes the self-excited oscillation of the local oscillator amplifier in pulsed mode. Preventive measures are given after analysis of the mechanism.Attach wave-absorbing materials to the sidewalls of the RF box can effectively eliminate self-excited oscillation.This article has value for the analysis and treatment of cavity self-excited oscillation in the future.

Abstract：Hall array current sensor is a research hotspot in the field of power electronics， especially in the measurement of direct current. It has the advantages of high linearity， large dynamic range， small size and low power consumption， which is an improvement of current sensor with ferromagnetic core. However， due to the absence of core poly-magnetic， in automotive electronics， distribution systems and other places， the conductors are densely arranged， and the external magnetic field is easy to enter the magnetic sensor， thus affecting the measurement accuracy. Therefore， this paper proposes an antimagnetic interference algorithm to solve this problem， using the method of constructing transcendental equations， which can calculate not only the measured current， but also the interference current at any position. The method is verified by numerical simulation， finite element analysis and laboratory experiments. The experimental results show that the error is less than 3.5% under the influence of strong interference current， which verifies the effectiveness of the circular sensor array and anti-magnetic interference algorithm.

Abstract：To overcome the challenges posed by traditional Partial Discharge (PD) fault diagnosis methods, such as their inability to effectively process complex, noisy PD signals and their reliance on manual denoising and expert input, and the difficulty on learning diverse PD feature representations, this paper introduces two advanced solutions: Residual Attention Adaptive Denoising Network (RAADNet) and Stacking ensemble-based PD fault diagnosis model. RAADNet leverages a residual network framework integrated with a Channel Attention Module (CAM) and a soft thresholding function for adaptive noise reduction. The Stacking ensemble model comprises distinct base learners, including the RAADNet with convolutional neural network architecture, a Transformer featuring multi-head self-attention, and an XGBoost model that adopts a Boosting strategy. Experimental findings reveal that RAADNet surpasses competing advanced techniques, achieving an accuracy of 93.99%. The Stacking model further improves performance by leveraging diverse feature representations, reaching an accuracy of 96.79%.

Abstract：Time Sensitive Network (TSN) combines the flexibility of traditional Ethernet “Best Effort” transmission methods with the real-time and deterministic characteristics of time sensitive networks, and has the ability to support different types of applications. The Per-Stream Filtering and Policing (PSFP) protocol of TSN can prevent illegal packet attacks and traffic overload by providing certain strategies. Based on the working principle of the protocol, the functions of stream filters, stream gates and flow metering are studied, and an implementation method of PSFP is proposed. Besides, the network topology is designed and verified through experiments on TSN switches. The results show that this design can successfully match abnormal packets that meet the rules, and block or limit the bandwidth of them according to requirements, which can prove the effectiveness of the design. This method is not limited to TSN implemented by hardware or software, and has certain reference value for the application of TSN in sectors such as industry and vehicle.

Abstract：To cope with new types of cyber attacks (e.g. false data injection attacks) that may be applied to smart grid systems, a machine learning-based intrusion detection method is proposed. The method employs an autoencoder for data dimensionality reduction and uses an extreme random tree classifier to detect potential attacks. The performance of the method is tested under different system sizes and attack levels based on IEEE standard power system data. The experimental results show that in the IEEE 118-node system, the detection accuracy of the method is as high as 99.76%, and the F1 score reaches 99.77% even when only 0.1% of the attack measurements are available, which is much higher than other algorithms. This method is not only effective in detecting intrusions in smart grids, but also has high computational efficiency.

Abstract：Regular expression matching technology plays a significant role in data processing tasks such as data cleaning, parsing, and extraction. However, due to issues such as strong data dependency and unpredictable memory access in the matching process, the matching performance is relatively low. In response to this problem, this paper proposes a high-performance regular expression data processing algorithm based on vector instruction set, which is called ßFA. By using vector instructions to read a sequence of consecutive characters at once, and performing vector matching with the non-trusted character set corresponding to the most frequently accessed state, built-in functions can be utilized to find the position of the first non-trusted character, thus obtaining the number of characters that can be skipped directly, thereby accelerating the matching performance. Experimental results show that the throughput of the ßFA algorithm is superior to the original DFA algorithm and the αFA algorithm, being 4.67~60 times faster than the original DFA algorithm and 4.37~7.82 times faster than the αFA algorithm.

Abstract：In the ongoing process of agricultural development, the health of crops continues to be a pivotal research area. Addressing this issue, this paper endeavors to delve into the classification of rice diseases within the framework of intelligent agricultural planting. Within the context of intelligent agricultural planting, this paper advocates the adoption of federated learning as a means to enhance the accuracy of disease classification equipment while safeguarding the data privacy of individual devices, thereby addressing the data silo problem among these devices. For the experimental phase, seven pre-trained models are meticulously selected to extract pertinent features, and four evaluation metrics—accuracy, recall, loss function, and F1-score—are employed to assess the performance of these models. The experimental outcomes reveal that the VGG19 model achieved remarkable accuracy levels of 99.05% and 98.48% on Independent and Identically Distributed (IID) and Non-Independent and Identically Distributed (Non-IID) data sets, respectively, showcasing its robustness and precision. Through a series of experiments and comparative analyses of various indicators, it is conclusively established that the integration of federated learning has enhanced the accuracy of the equipment by a noteworthy margin of 4.36%. Furthermore, the convergence time of the image classification model is influenced by a combination of factors, including the number of federated learning rounds and the training epochs per round within the training set. Notably, the stability of the model improves as the number of devices participating in federated learning increases.

Abstract：In order to improve the security of Internet of Vehicles（IoV）, a network security enhancement framework for mobile edge computing based on neuro-fuzzy system and blockchain is proposed. The framework combines technologies such as bulk authentication and key exchange, combines blockchain with mobile edge computing, and is architecturally divided into three layers: perception layer, edge computing layer, and service layer.The perception layer ensures data security through blockchain, the edge computing layer provides resources, and the service layer protects the data using blockchain technology. Simulation shows that combining blockchain and mobile edge computing can effectively improve the security of IoV.

Abstract：Electromechanical control system contains many kinds of chips, and its miniaturization and lightweight demand is increasingly urgent. As an advanced packaging method, system level packaging （System in Package）technology can integrate many different types of chips in a smaller space. In this paper, a SiP circuit of electromechanical control is designed based on system-level packaging technology, combined with TSV and FanOUT technology. The circuit includes the DSP signal control unit and the bottom FPGA signal processing unit, which are stacked in the form of PoP to form a SiP circuit. Compared with the electromechanical control system built by conventional discrete devices, the SiP volume is reduced by more than 70% and the weight is reduced by more than 80%. A test system is designed for SiP circuit, and a loopback test method is proposed for ADC and DAC chips, which can improve the test efficiency. The results show that the circuit meets the design requirements and has a certain application prospect in the field of electromechanical control.

Abstract：This article proposes a design method for a wide input voltage and high efficiency 100 W space secondary power supply. The input voltage range reaches 20 V~50 V, which can cover the voltage range of 28 V and 42 V buses on the satellite. The peak efficiency reaches over 90%, with the characteristics of small size and high power density. Adopting a single ended flyback circuit topology and incorporating slope compensation technology, the duty cycle of the flyback converter operates at over 60%, expanding the input voltage range. Simultaneously using planar transformer technology to reduce transformer leakage inductance, improve efficiency, and reduce size.

Abstract：In this paper, the location problem of intermodulation interference in satellite communication system is studied. The mechanism of intermodulation interference is analyzed, and the location method of intermodulation interference is given. Based on the quantitative analysis of intermodulation interference spectrum characteristics and the frequency allocation of each station by spectrum control station, a dynamic database of each station frequency and frequency difference in the system is established. An intermodulation interference source location method based on frequency difference is designed to accurately locate the interference stations. The simulation results verify the feasibility of the proposed scheme in the intermodulation interference positioning process, effectively reduce the time of disturbance positioning and improve the probability of accurate positioning of the disturbance station, which provides a solution for the rapid positioning of the satellite communication system in the face of such interference, and has great practical significance for the stable operation of the satellite communication network.