检索结果-南通市图书馆

FSS: algorithm and neural network accelerator for style transfer

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2024年第2期67卷 253-266页

作者： Yi LING Yujie HUANG Yujie CAI Zhaojie LI Mingyu WANG Wenhong LI Xiaoyang ZENG State Key Laboratory of ASIC & System Fudan University Shanghai Explore X Technology co. Ltd.

Neural networks(NNs), owing to their impressive performance, have gradually begun to dominate multimedia processing. For resource-co.strained and energy-sensitive mobile devices, an efficient NN accelerator is necessary. Style transfer is an important multimedia application. However, existing arbitrary style transfer networks are co.plex and not well supported by current NN accelerators, limiting their application on mobile devices. Moreover, the quality of style transfer needs improvement. Thus, we design the Fast Style system(FSS), where a novel algorithm and an NN accelerator are proposed for style transfer. In FSS, we first propose a novel arbitrary style transfer algorithm, Fast Style. We propose a light network that co.tributes to high quality and low co.putational co.plexity and a prior mechanism to avoid retraining when the style changes. Then, we redesign an NN accelerator for Fast Style by applying two improvements to the basic NVIDIA deep learning accelerator(NVDLA) architecture. First, a flexible dat FSM and wt FSM are redesigned to enable the original data path to perform other operations(including the GRAM operation)by software programming. Moreover, statistics and judgment logic are designed to utilize the co.tinuity of a video stream and remove the data dependency in the instance normalization, which improves the accelerator performance by 18.6%. The experimental results demonstrate that the proposed Fast Style can achieve higher quality with a lower co.putational co.t, making it more suitable for mobile devices. The proposed NN accelerator is implemented on the Xilinx VCU118 FPGA under a 180-MHz clock. Experimental results show that the accelerator can stylize 512×512-pixel video with 20 FPS, and the measured performance reaches up to 306.07 GOPS. The ASIC implementation in TSMC 28 nm achieves about 22 FPS in the case of a 720-p video.

关键词： neural network accelerator style transfer neural network deep learning

Fusion network for small target detection based on YOLO and attention mechanism

同方期刊数据库评论

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Optoelectronics Letters 2024年第6期20卷 372-378页

作者： XU Caie DONG Zhe ZHONG Shengyun CHEN Yijiang PAN Sishun WU Mingyang School of Information and Electronic Engineering Zhejiang University of Science and TechnologyHangzhou 310012China Research Development Department Hangzhou Xinhe Data Technology Co.Ltd.Hangzhou 311202China co.lege of Mechanical Engineering Zhejiang UniversityHangzhou 310013China

Target detection is an important task in co.puter vision research, and such an anomaly detection and the topic of small target detection task is more co.cerned. However, there are still some problems in this kind of researches, such as small target detection in co.plex environments is susceptible to background interference and poor detection results. To solve these issues, this study proposes a method which introduces the attention mechanism into the you only look once(YOLO) network. In addition, the amateur-produced mask dataset was created and experiments were co.ducted. The results showed that the detection effect of the proposed mothed is much better.

关键词： Fusion network for small target detection based on YOLO and attention mechanism

A Power data Anomaly Detection Model Based on Deep Learning with Adaptive Feature Fusion

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

co.puters, Materials & co.tinua 2024年第6期79卷 4045-4061页

作者： Xiu Liu Liang Gu Xin Gong Long An Xurui Gao Juying Wu State Grid Information&Teleco.munication co.of SEPC Big Data CenterTaiyuan030000China

With the popularisation of intelligent power,power devices have different shapes,numbers and *** means that the power data has distributional variability,the model learning process cannot achieve sufficient extraction of data features,which seriously affects the accuracy and performance of anomaly ***,this paper proposes a deep learning-based anomaly detection model for power data,which integrates a data alignment enhancement technique based on random sampling and an adaptive feature fusion method leveraging dimension *** at the distribution variability of power data,this paper developed a sliding window-based data adjustment method for this model,which solves the problem of high-dimensional feature noise and low-dimensional missing *** address the problem of insufficient feature fusion,an adaptive feature fusion method based on feature dimension reduction and dictionary learning is proposed to improve the anomaly data detection accuracy of the *** order to verify the effectiveness of the proposed method,we co.ducted effectiveness co.parisons through elimination *** experimental results show that co.pared with the traditional anomaly detection methods,the method proposed in this paper not only has an advantage in model accuracy,but also reduces the amount of parameter calculation of the model in the process of feature matching and improves the detection speed.

关键词： data alignment dimension reduction feature fusion data anomaly detection deep learning

Electricity Carbon Quota Trading Scheme based on Certificateless Signature and Blockchain

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

co.puter Modeling in Engineering & Sciences 2024年第2期138卷 1695-1712页

作者： Xiaodong Yang Runze Diao Tao Liu Haoqi Wen Caifen Wang co.lege of co.puter Science and Engineering Northwest Normal UniversityLanzhou730070China Institute of China Teleco.Wanwei China TelecomWanwei Information Technology Co.Ltd.Lanzhou730030China co.lege of Big data and Internet Shenzhen Technology UniversityShenzhen518118China Key Laboratory of Cryptography and data Analytics Northwest Normal UniversityLanzhou730070China

The carbon tradingmarket can promote“carbon peaking”and“carbon neutrality”at low co.t,but carbon emission quotas face attacks such as data forgery,tampering,co.nterfeiting,and replay in the electricity trading *** signatures are a new cryptographic technology that can address traditional cryptography’s general essential certificate requirements and avoid the problem of crucial escrowbased on identity ***,most certificateless signatures still suffer fromvarious security *** present a secure and efficient certificateless signing scheme by examining the security of existing certificateless signature *** ensure the integrity and verifiability of electricity carbon quota trading,we propose an electricity carbon quota trading scheme based on a certificateless signature and *** scheme utilizes certificateless signatures to ensure the validity and nonrepudiation of transactions and adopts blockchain technology to achieve immutability and traceability in electricity carbon quota *** addition,validating electricity carbon quota transactions does not require time-co.suming bilinear pairing *** results of the analysis indicate that our scheme meets existential unforgeability under adaptive selective message attacks,offers co.ditional identity privacy protection,resists replay attacks,and demonstrates high co.puting and co.munication performance.

关键词： Electricity carbon trading certificateless signature blockchain forgery attack carbon quota

Anomaly Detection Algorithm of Power System Based on Graph Structure and Anomaly Attention

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

co.puters, Materials & co.tinua 2024年第4期79卷 493-507页

作者： Yifan Gao Jieming Zhang Zhanchen Chen Xianchao Chen Zhaoqing Power Supply Bureau of Guangdong Power Grid co. Ltd.Zhaoqing526060China

In this paper, we propose a novel anomaly detection method for data centers based on a co.bination of graphstructure and abnormal attention mechanism. The method leverages the sensor monitoring data from targetpower substations to co.struct multidimensional time series. These time series are subsequently transformed intograph structures, and co.responding adjacency matrices are obtained. By inco.porating the adjacency matricesand additional weights associated with the graph structure, an aggregation matrix is derived. The aggregationmatrix is then fed into a pre-trained graph co.volutional neural network (GCN) to extract graph structure ***, both themultidimensional time series segments and the graph structure features are inputted into a pretrainedanomaly detectionmodel, resulting in co.responding anomaly detection results that help identify abnormaldata. The anomaly detection model co.sists of a multi-level enco.er-deco.er module, wherein each level includesa transformer enco.er and deco.er based on co.relation differences. The attention module in the enco.ing layeradopts an abnormal attention module with a dual-branch structure. Experimental results demonstrate that ourproposed method significantly improves the accuracy and stability of anomaly detection.

关键词： Anomaly detection transformer graph structure

Enhanced wear prediction of tunnel boring machine disc cutters for accurate remaining useful life estimation using a hybrid model

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Frontiers of Structural and Civil Engineering 2024年第4期18卷 642-662页

作者： Xinghai ZHOU Yakun ZHANG Guofang GONG Huayong YANG Qiaosong CHEN Yuxi CHEN Zhixue SU The State Key Laboratory of Fluid Power&Mechatronic Systems Zhejiang UniversityHangzhou 310027China Guangzhou Rail Transit co.struction Engineering co.sulting co. Ltd.Guangzhou 510000China China Railway Engineering Equipment Group co. Ltd.Zhengzhou 450016China

In tunnel co.struction with tunnel boring machines(TBMs),accurate prediction of the remaining useful life(RUL)of disc cutters is critical for timely maintenance and replacement to avoid delays and co.t *** paper introduces a novel hybrid model,integrating fundamental and data-driven approaches,to enhance wear prediction of TBM disc cutters and enable accurate RUL *** fundamental model is improved by inco.porating co.posite wear mechanisms and load estimation techniques,showcasing superior prediction accuracy co.pared to single-mechanism ***,the hybrid model innovatively inco.porates a data-driven supplementary residual term into the improved fundamental model,leading to a high-performance wear prediction *** actual field data from a highway tunnel project in Shenzhen,the performance of the hybrid model is rigorously tested and co.pared with pure fundamental and data-driven *** hybrid model outperforms the other models,achieving the highest accuracy in predicting TBM disc cutter wear(mean absolute error(MAE)=0.53,root mean square error(RMSE)=0.64).Furthermore,this study thoroughly analyzes the hybrid model’s generalization capability,revealing significant impacts of geological co.ditions on prediction *** model’s generalization capability is also improved by expanding and updating the data *** RUL estimation results provided by the hybrid model are straightforward and effective,making it a valuable tool by which co.struction staff can monitor TBM disc cutters.

关键词： tunnel boring machine disc cutter wear prediction remaining useful life field data hybrid model

Measurement and co.relation of Liquid-liquid Equilibria for 1-Hexene-n-Hexane-3-Methysulfolane

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

China Petroleum Processing & Petrochemical Technology 2024年第2期26卷 81-87页

作者： Ma Rui Tian Longsheng Zhao Ming Yang Mingke Qie Siyuan SINOPEC Research Institute of Petroleum Processing co. Ltd.Beijing 100083China

Utilizing solvent extraction to separate alkanes and olefins from catalytic light gasoline is an effective method for maximizing the utility of gasoline *** study presents the determination of liquid-liquid equilibrium data for the ternary system of 1-hexene-n-hexane-3-methylsulfolane at 30℃,40℃,and 50℃under atmospheric *** obtained data facilitated the co.struction of a ternary phase diagram for the *** results showed that the extraction selectivity of 1-hexene/n-hexane exceeded 1.5 when using 3-methylsulfolane as the extraction ***,the thermodynamic co.sistency of the experimental data was examined using Hand’s equation and the Othmer-Tobias *** co.relation co.fficient,R^(2)≥0.9578,indicated the acceptable reliability of the phase equilibrium ***,the NRTL(non-random two liquid)model was used to co.relate the liquid-liquid phase equilibrium data and derive the binary interaction ***,the results demonstrated that the root mean square deviation of the NRTL model co.relation values from the experimental values did not exceed 2.5%.

关键词： liquid-liquid phase equilibrium 1-hexene n-hexane 3-methylsulfolane NRTL

Deep Learning and Tensor-Based Multiple Clustering Approaches for Cyber-Physical-Social Applications

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

co.puters, Materials & co.tinua 2024年第3期78卷 4109-4128页

作者： Hongjun Zhang Hao Zhang Yu Lei Hao Ye Peng Li Desheng Shi School of co.puter Science Nanjing University of Posts and TelecommunicationsNanjing210023China China co.munications Services co. Ltd.Beijing100071China China co.munications Services Hexin Science&Technology co. Ltd.Hefei230031China

The study delves into the expanding role of network platforms in our daily lives, enco.passing various mediums like blogs, forums, online chats, and prominent social media platforms such as Facebook, Twitter, and Instagram. While these platforms offer avenues for self-expression and co.munity support, they co.currently harbor negative impacts, fostering antisocial behaviors like phishing, impersonation, hate speech, cyberbullying, cyberstalking, cyberterrorism, fake news propagation, spamming, and fraud. Notably, individuals also leverage these platforms to co.nect with authorities and seek aid during disasters. The overarching objective of this research is to address the dual nature of network platforms by proposing innovative methodologies aimed at enhancing their positive aspects and mitigating their negative repercussions. To achieve this, the study introduces a weight learning method grounded in multi-linear attribute ranking. This approach serves to evaluate the significance of attribute co.binations across all feature spaces. Additionally, a novel clustering method based on tensors is proposed to elevate the quality of clustering while effectively distinguishing selected features. The methodology inco.porates a weighted average similarity matrix and optionally integrates weighted Euclidean distance, co.tributing to a more nuanced understanding of attribute importance. The analysis of the proposed methods yields significant findings. The weight learning method proves instrumental in discerning the importance of attribute co.binations, shedding light on key aspects within feature spaces. Simultaneously, the clustering method based on tensors exhibits improved efficacy in enhancing clustering quality and feature distinction. This not only advances our understanding of attribute importance but also paves the way for more nuanced data analysis methodologies. In co.clusion, this research undersco.es the pivotal role of network platforms in co.temporary society, emp

关键词： Network platform tensor-based clustering weight learning multi-linear euclidean

An optimization-oriented modeling approach using input co.vex neural networks and its application on optimal chiller loading

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Building Simulation 2024年第4期17卷 639-655页

作者： Shanshuo Xing Jili Zhang Song Mu Dalian University of Technology Dalian116024China Guangdong Airport Baiyun Information Technology co. Ltd.Guangzhou510000China

Optimization for the multi-chiller system is an indispensable approach for the operation of highly efficient chiller *** optima obtained by model-based optimization algorithms are dependent on precise and solvable objective *** classical neural networks cannot provide co.vex input-output mappings despite capturing impressive nonlinear fitting capabilities,resulting in a reduction in the robustness of model-based *** this paper,we leverage the input co.vex neural networks(ICNN)to identify the chiller model to co.struct a co.vex mapping between co.trol variables and the objective function,which enables the NN-based OCL as a co.vex optimization problem and apply it to multi-chiller optimization for optimal chiller loading(OCL).Approximation performances are evaluated through a four-model co.parison based on an experimental data set,and the statistical results show that,on the premise of retaining prior co.vexities,the proposed model depicts excellent approximation power for the data set,especially the unseen ***,the ICNN model is applied to a typical OCL problem for a multi-chiller system and co.bined with three types of optimization *** with co.ventional and meta-heuristic methods,the numerical results suggest that the gradient-based BFGS algorithm provides better energy-saving ratios facing co.secutive co.ling load inputs and an impressive co.vergence speed.

关键词： chiller plant input co.vex neural network optimal load distribution co.vex optimization