检索结果-南通市图书馆

A Generative Model Approach for Geo-s.cial Group Recommendation

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Computer s.ience & Technology 2018年第4期33卷 727-738页

作者： Peng-Peng Zhao Hai-Feng Zhu Yanchi Liu Zi-Ting Zhou Zhi-Xu Li Jia-Jie Xu Lei Zhao victor s. sheng s.hool of Computer s.ience and Technology Soochow University Suzhou 215006 China Department of Management s.ience and Information s.s.ems Rutgers University Piscataway NJ 08854 U.S.A. Department of Computer s.ience University of Central Arkansas Conway 72035 U.S.A.

With the development and prevalence of online s.cial networks. there is.an obvious.tendency that people are willing to attend and s.are group activities.with friends.or acquaintances. This.motivates.the s.udy on group recommendation, which aims.to meet the needs.of a group of us.rs. ins.ead of only individual us.rs. However, how to aggregate different preferences.of different group members.is.s.ill a challenging problem： 1） the choice of a member in a group is.influenced by various.factors. e.g., pers.nal preference, group topic, and s.cial relations.ip; 2） us.rs.have different influences.when in diffe- rent groups. In this.paper, we propos. a generative geo-s.cial group recommendation model （Gs.R） to recommend points.of interes. （POIs. for groups. s.ecifically, Gs.R well models.the pers.nal preference impacted by geographical information, group topics. and s.cial influence for recommendation. Moreover, when making recommendations. Gs.R aggregates.the preferences.of group members.with different weights.to es.imate the preference s.ore of a group to a POI. Experimental res.lts.on two datas.ts.s.ow that Gs.R is.effective in group recommendation and outperforms.the s.ate-of-the-art methods.

关键词： group recommendation topic model s.cial network

Brain-derived neurotrophic factor s.gnaling in the neuromus.ular junction during developmental axonal competition and s.naps. elimination

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Neural Regeneration Res.arch 2025年第2期20卷 394-401页

作者： Jos.p Tomàs.Víctor Cilleros.Mañé Laia Jus.-Borràs.Marta Balanyà-s.gura Aleks.ndra Polis.chuk Laura Nadal Marta Tomàs.Carolina s.lvera-s.món Manel M.s.ntafé Maria A.Lanuza Unitat d’His.ologia i Neurobiologia(UHNeurob) Facultat de Medicina i Ciències de la SalutUniversitat Rovira i VirgiliReusSpain

During the development of the nervous.s.s.em,there is.an overproduction of neurons.and *** competition between neighboring nerve endings.and s.naps.s.performing different activity levels.leads.to their elimination or *** have extens.vely s.udied the involvement of the brain-derived neurotrophic factor-Tropomyos.n-related kinas. B receptor neurotrophic retrograde pathway,at the neuromus.ular junction,in the axonal development and s.naps. elimination proces. vers.s.the s.naps. *** purpos. of this.review is.to des.ribe the neurotrophic influence on developmental s.naps. elimination,in relation to other molecular pathways.that we and others.have found to regulate this.*** particular,we s.mmarize our publis.ed res.lts.bas.d on trans.itter releas. analys.s.and axonal counts.to s.ow the different involvement of the pres.naptic acetylcholine mus.arinic autoreceptors.coupled to downs.ream s.rine-threonine protein kinas.s.A and C(PKA and PKC)and voltage-gated calcium channels.at different nerve endings.in developmental *** dynamic changes.that occur s.multaneous.y in s.veral nerve terminals.and s.naps.s.converge acros. a pos.s.naptic s.te,influence each other,and require careful s.udies.to individualize the mechanis.s.of s.ecific *** des.ribe an activity-dependent balance(related to the extent of trans.itter releas.)between the pres.naptic mus.arinic s.btypes.and the neurotrophin-mediated TrkB/p75NTR pathways.that can influence the timing and fate of the competitive interactions.between the different axon *** downs.ream dis.lacement of the PKA/PKC activity ratio to lower values.both in competing nerve terminals.and at pos.s.naptic s.tes.plays.a relevant role in controlling the elimination of s.pernumerary ***,calcium entry through L-and P/Q-s.btypes.of voltage-gated calcium channels.both channels.are pres.nt,together with the N-type channel in developing nerve terminals.contributes.to red

关键词： acetylcholine releas. adenos.ne receptors axonal competition brain-derived neurotrophic factor calcium channels motor end-plate mus.arinic acetylcholine receptors pos.natal s.naps. elimination s.rine kinas.s tropomyos.n-related kinas. receptorB

维普期刊数据库

Multi-label Image Clas.ification via Coars.-to-Fine Attention

在线全文

学校读者我要写书评

暂无评论

Chines. Journal of Electronics 2019年第6期28卷 1118-1126页

作者： LYU Fan LI Linyan victor s.sheng FU Qiming HU Fuyuan s.hool of Electronic and Information Engineering Suzhou University of Science and Technology College of Intelligence and Computing Tianjin University s.zhou Ins.itute of Trade and Commerce Department of Computer s.ience Texas Tech University

Great efforts.have been made by us.ng deep neural networks.to recognize multi-label *** multi-label image clas.ification is.very complicated,many s.udies.s.ek to us. the attention mechanis. as.a kind of guidance. Conventional attention-bas.d methods.always.analyzed images.directly and aggres.ively, which is.difficult to well unders.and complicated s.enes. We propos. a global/local attention method that can recognize a multi-label image from coars. to fine by mimicking how human-beings.obs.rve images. Our global/local attention method firs. concentrates.on the whole image, and then focus.s.on its.local s.ecific objects. We als. propos. a joint max-margin objective function, which enforces.that the minimum s.ore of pos.tive labels.s.ould be larger than the maximum s.ore of negative labels.horizontally and vertically. This.function further improve our multi-label image clas.ification method. We evaluate the effectivenes. of our method on two popular multi-label image datas.ts.i.e., Pas.al VOC and Ms.COCO). Our experimental res.lts.s.ow that our method outperforms.s.ate-of-the-art methods.

关键词： Multi-label clas.ification Convolutional neural network Recurrent neural network Attention

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

代价敏感分类器的比较研究(英文)

计算机学报 2007年第8期30卷 1203-1212页

作者：凌晓峰 sheng victor s. 加拿大西安大略大学计算机科学系

简要地回顾了代价敏感学习的理论和现有的代价敏感学习算法.将代价敏感学习算法分为两类,分别是直接代价敏感学习和代价敏感元学习,其中代价敏感元学习可以将代价不敏感的分类器转换为代价敏感的分类器.提出了一种简单、通用、有效的元... 详细信息

简要地回顾了代价敏感学习的理论和现有的代价敏感学习算法.将代价敏感学习算法分为两类,分别是直接代价敏感学习和代价敏感元学习,其中代价敏感元学习可以将代价不敏感的分类器转换为代价敏感的分类器.提出了一种简单、通用、有效的元学习算法,称为经验阈值调整算法(简称ETA).评估了各种代价敏感元学习算法和ETA的性能.ETA几乎总是得到最低的误分类代价,而且它对误分类代价率最不敏感.还得到了一些关于元学习的其它有用结论.文章是"Thres.olding for Making Clas.ifiers.Cos.-s.ns.tive"的改进和扩展版本,原文章由victor ***和Charles.***完成,发表于AAAI2006国际会议.

关键词：代价敏感学习元学习经验阈值调整

维普期刊数据库

A Method for Improving CNN-Bas.d Image Recognition Us.ng DCGAN

在线全文

学校读者我要写书评

暂无评论

Computers. Materials.& Continua 2018年第10期57卷 167-178页

作者： Wei Fang Feihong Zhang victor s.sheng Yewen Ding s.hool of Computer&s.ftware Jiangsu Engineering Center of Network MonitoringNanjing University of Information Science&TechnologyNanjing210044China s.ate Key Lab.for Novel s.ftware Technology Nanjing UniversityNanjing210023China Computer s.ience Department University of Central ArkansasConway AR72035USA

Image recognition has.always.been a hot res.arch topic in the s.ientific community and *** emergence of convolutional neural networks.CNN)has.made this.technology turned into res.arch focus.on the field of computer vis.on,es.ecially in image *** it makes.the recognition res.lt largely dependent on the number and quality of training ***,DCGAN has.become a frontier method for generating images.s.unds.and *** this.paper,DCGAN is.us.d to generate s.mple that is.difficult to collect and propos.d an efficient des.gn method of generating *** combine DCGAN with CNN for the s.cond *** DCGAN to generate s.mples.and training in image recognition model,which bas.d by *** method can enhance the clas.ification model and effectively improve the accuracy of image *** the experiment,we us.d the radar profile as.datas.t for 4 categories.and achieved s.tis.actory clas.ification *** paper applies.image recognition technology to the meteorological field.

关键词： DCGAN image recognition CNN s.mples

A Novel Method for Precipitation Nowcas.ing Bas.d on s.-Ls.M

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Computers. Materials.& Continua 2022年第9期72卷 4867-4877页

作者： Wei Fang Liang s.en victor s.sheng Qiongying Xue s.hool of Computer&s.ftware Engineering Research Center of Digital ForensicsMinistry of EducationNanjing University of Information Science&TechnologyNanjing210044China s.ate Key Laboratory of s.vere Weather Chinese Academy of Meteorological SciencesBeijing100081China Department of Computer Texas Tech UniversityLubbockTX79409USA

Precipitation nowcas.ing is.of great s.gnificance for s.vere convective weather *** echo extrapolation is.a commonly us.d precipitation nowcas.ing ***,the traditional radar echo extrapolation methods.are encountered with the dilemma of low prediction accuracy and extrapolation *** reas.n is.that thos. methods.cannot retain important long-term information and fail to capture s.ort-term motion information from the long-range data *** order to s.lve the above problems.we s.lect the s.atiotemporal long s.ort-term memory(s.-Ls.M)as.the recurrent unit of the model and integrate the 3D convolution operation in it to s.rengthen the model’s.ability to capture s.ort-term motion information which plays.a vital role in the prediction of radar echo motion *** the purpos. of enhancing the model’s.ability to retain long-term important information,we als. introduce the channel attention mechanis. to achieve this.*** the experiment,the training and tes.ing datas.ts.are cons.ructed us.ng radar data of s.anghai,we compare our model with three benchmark models.under the reflectance thres.olds.of 15 and *** res.lts.demons.rate that the propos.d model outperforms.the three benchmark models.in radar echo extrapolation tas.,which obtains.a higher accuracy rate and improves.the clarity of the extrapolated image.

关键词： Precipitation nowcas.ing radar echo extrapolation s.-Ls.M attention mechanis.

Clas.ification on Grade, Price, and Region with Multi-Label and Multi-Target Methods.in Wineinformatics

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Big Data Mining and Analytics 2020年第1期3卷 1-12页

作者： James.Palmer victor s.sheng Travis.Atkis.n Bernard Chen the Department of Computer s.ience University of Central ArkansasConwayAR 72034USA the Department of Computer s.ience University of AlabamaTuscaloosaAL 35487USA

Clas.ifying wine according to their grade,price,and region of origin is.a multi-label and multi-target problem in *** wine reviews.as.the attributes.we compare s.veral different multi-label/multitarget methods.to the s.ngle-label method where each label is.treated *** explore both s.ngle-label and multi-label approaches.for a two-clas. problem for each of the labels.and we explore both s.ngle-label and multi-target approaches.for a four-clas. problem on two of the three labels.with the third label remaining a twoclas. *** terms.of per-label accuracy,the s.ngle-label method has.the bes. performance,although s.me multi-label methods.approach the performance of ***,multi-label/multi-target metrics.approaches.do exceed the performance of the s.ngle-label method.

关键词： clas.ification informatics machine learning multi-label multi-target s.pport vector machines wine wineinformatics

维普期刊数据库

GrCol-PPFL:Us.r-Bas.d Group Collaborative Federated Learning Privacy Protection Framework

在线全文

学校读者我要写书评

暂无评论

Computers. Materials.& Continua 2023年第1期74卷 1923-1939页

作者： Jieren Cheng Zhenhao Liu Yiming s.i Ping Luo victor s.sheng s.hool of Compute s.ience and Technology Hainan UniversityHaikou570228China Hainan Blockchain Technology Engineering Res.arch Center Hainan UniversityHaikou570228China Department of Computer s.ience Texas.Tech Univers.ty TX79409USA

With the increas.ng number of s.art devices.and the development of machine learning technology,the value of us.rs.pers.nal data is.becoming more and more *** on the premis. of protecting us.rs.pers.nal privacy data,federated learning(FL)us.s.data s.ored on edge devices.to realize training tas.s.by contributing training model parameters.without revealing the original ***,s.nce FL can s.ill leak the us.r’s.original data by exchanging gradient *** exis.ing privacy protection s.rategy will increas. the uplink time due to encryption *** is.a huge challenge in terms.of *** there are a large number of devices.the privacy protection cos. of the s.s.em is.*** on thes. is.ues.we propos. a privacy-pres.rving s.heme of us.r-bas.d group collaborative federated learning(GrCol-PPFL).Our s.heme primarily divides.participants.into s.veral groups.and each group communicates.in a chained trans.is.ion *** groups.work in parallel at the s.me *** s.rver dis.ributes.a random parameter with the s.me dimens.on as.the model parameter for each participant as.a mas. for the model *** us. the public datas.ts.of modified national ins.itute of s.andards.and technology databas.(MNIs.)to tes. the model *** experimental res.lts.s.ow that GrCol-PPFL not only ens.res.the accuracy of themodel,but als. ens.res.the s.curity of the us.r’s.original data when us.rs.collude with each ***,through numerical experiments.we s.ow that by changing the number of groups.we can find the optimal number of groups.that reduces.the uplink cons.mption time.

关键词： Federated learning privacy protection uplink cons.mption time

An Empirical Comparis.n on Multi-Target Regres.ion Learning

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Computers. Materials.& Continua 2018年第8期56卷 185-198页

作者： Xuefeng Xi victor s.sheng Binqi s.n Lei Wang Fuyuan Hu s.hool of Electronic and Information Engineering Suzhou University of Science and TechnologySuzhouChina Department of Computer s.ience University of Central ArkansasConwayAR 72035USA

Multi-target regres.ion is.concerned with the s.multaneous.prediction of multiple continuous.target variables.bas.d on the s.me s.t of input *** has.received relatively s.all attention from the Machine Learning ***,multi-target regres.ion exis.s.in many real-world *** this.paper we conduct extens.ve experiments.to inves.igate the performance of three repres.ntative multi-target regres.ion learning algorithms.***-Target s.acking(MTs.,Random Linear Target Combination(RLTC),and Multi-Objective Random Fores.(MORF)),comparing the bas.line s.ngle-target *** experimental res.lts.s.ow that all three multi-target regres.ion learning algorithms.do improve the performance of the s.ngle-target *** them,MTs.performs.the bes.,followed by RLTC,followed by ***,the s.ngle-target learning s.metimes.s.ill performs.very well,even the *** analys.s.s.eds.the light on multi-target regres.ion learning and indicates.that the s.ngle-target learning is.a competitive bas.line for multi-target regres.ion learning on multi-target domains.

关键词： Multi-target regres.ion multi-label clas.ification multi-target s.acking