新闻频道 > 新政风向

Content to Node: Self-Translation Network Embedding_骑士火眼晶晶资讯网

来源: 新华社
01:05:35

江苏气象官员落马_Content to Node: Self-Translation Network Embedding

paper:https://dl.acm.org/citation.cfm?id=3219988
data & code:http://dm.nankai.edu.cn/code/STNE.rar
?

简介

目的:学习网络中节点的低维表示
将利用节点信息及网络结构识别节点的过程比喻为翻译的过程
?
论文引用网络中,每个节点代表一篇论文,每条边代表引用关系。每个节点自身属性包括文章的摘要,关键词,研究领域等等。该论文的假设依据是,论文所形成的引用网络与论文自身的属性之间有较强关系
?
现有方案
1. 将结构信息和属性信息分别进行embedding之后,组合
2. 考虑短距离/固定邻域范围保留结构信息(第一/二邻近)
(复杂问题中很难确定邻域范围)
?

创新点(贡献):

提出基于seq2seq 的模型框架(STNE)
利用网络上随机遍历生成的序列,将节点内容信息翻译成结构信息,从而结合两种信息
?
  • 将网络嵌入转化为 seq2seq 任务,从局部建模到序列的全局结构建模,捕获更多语义信息
  • 设计了一个异构的seq2seq 模型,嵌入原始输入文本,以端到端的方式学习从节点属性序列到节点指示序列的映射
?
与传统方法相比, STNE 直接对建模节点序列,从文本序列中自动学习生成函数,将 seq2seq 网络模型与其他文本嵌入模型相结合,通过学习内容序列到节点序列的映射,将内容信息和结构信息无缝融合到隐藏层的潜在向量中,能够高效表示节点
?
相对于 CANE: 从相邻文本节点感知嵌入
STNE:针对不同序列学习动态的节点嵌入(需要更长范围,更灵活的上下文)
?
框架
STNE 总体框架
?
该框架步骤:
1. 给定内容丰富的网络,通过随机游走提取节点序列,并将节点序列分为两个部分
节点属性序列
节点指示序列(由节点指示向量 one-hot 表示)
2. 通过这两个序列学习特定的 seq2seq 模型,该模型可以用于将节点属性“翻译”为节点指示向量
3. 步骤二的目的是得到中间层的潜在转换( 是可用于复杂网络分析 )
?
?更详细的总结
Self-Translation Network Embedding 阅读笔记

paper: https: dl. acm. org citation. cfm? id 3219988 data amp code: http: dm. nankai. edu. cn code STNE. rar? jian jie mu di: xue xi wang luo zhong jie dian de di wei biao shi jiang li yong jie dian xin xi ji wang luo jie gou shi bie jie dian de guo cheng bi yu wei fan yi de guo cheng? lun wen yin yong wang luo zhong, mei ge jie dian dai biao yi pian lun wen, mei tiao bian dai biao yin yong guan xi. mei ge jie dian zi shen shu xing bao kuo wen zhang de zhai yao, guan jian ci, yan jiu ling yu deng deng. gai lun wen de jia she yi ju shi, lun wen suo xing cheng de yin yong wang luo yu lun wen zi shen de shu xing zhi jian you jiao qiang guan xi? xian you fang an: 1. jiang jie gou xin xi he shu xing xin xi fen bie jin xing embedding zhi hou, zu he 2. kao lv duan ju li gu ding lin yu fan wei bao liu jie gou xin xi di yi er lin jin fu za wen ti zhong hen nan que ding lin yu fan wei? chuang xin dian gong xian: ti chu ji yu seq2seq de mo xing kuang jia STNE li yong wang luo shang sui ji bian li sheng cheng de xu lie, jiang jie dian nei rong xin xi fan yi cheng jie gou xin xi, cong er jie he liang zhong xin xi? jiang wang luo qian ru zhuan hua wei seq2seq ren wu, cong ju bu jian mo dao xu lie de quan ju jie gou jian mo, bu huo geng duo yu yi xin xi she ji le yi ge yi gou de seq2seq mo xing, qian ru yuan shi shu ru wen ben, yi duan dao duan di fang shi xue xi cong jie dian shu xing xu lie dao jie dian zhi shi xu lie de ying she? yu chuan tong fang fa xiang bi, STNE zhi jie dui jian mo jie dian xu lie, cong wen ben xu lie zhong zi dong xue xi sheng cheng han shu, jiang seq2seq wang luo mo xing yu qi ta wen ben qian ru mo xing xiang jie he, tong guo xue xi nei rong xu lie dao jie dian xu lie de ying she, jiang nei rong xin xi he jie gou xin xi wu feng rong he dao yin cang ceng de qian zai xiang liang zhong, neng gou gao xiao biao shi jie dian? xiang dui yu CANE: cong xiang lin wen ben jie dian gan zhi qian ru STNE: zhen dui bu tong xu lie xue xi dong tai de jie dian qian ru xu yao geng zhang fan wei, geng ling huo de shang xia wen? kuang jia STNE zong ti kuang jia? gai kuang jia bu zhou: 1. gei ding nei rong feng fu de wang luo, tong guo sui ji you zou ti qu jie dian xu lie, bing jiang jie dian xu lie fen wei liang ge bu fen jie dian shu xing xu lie jie dian zhi shi xu lie you jie dian zhi shi xiang liang onehot biao shi 2. tong guo zhe liang ge xu lie xue xi te ding de seq2seq mo xing, gai mo xing ke yi yong yu jiang jie dian shu xing" fan yi" wei jie dian zhi shi xiang liang 3. bu zhou er de mu di shi de dao zhong jian ceng de qian zai zhuan huan shi ke yong yu fu za wang luo fen xi ?? geng xiang xi de zong jie SelfTranslation Network Embedding yue du bi ji

当前文章:http://www.kiligar.com/mz68u5xh/693516-760577-56311.html

发布时间:04:12:40

139kj手机开奖现场??www.hh66.com??31367天机??天下彩天空彩正版资料??080355.com??王中王今晚开特结果??惠泽社群平码二中二??香港财神挂牌彩图??三怪玄机图??六合现场开奖结果??

关键词:鹰潭僚押科技,海军护航编队凯旋,武汉第纱传媒广告有限公司责任编辑:徒通