biane bipartite attributed network embedding
play

BiANE: Bipartite Attributed Network Embedding 1 2 2 1 3 Wentao - PowerPoint PPT Presentation

BiANE: Bipartite Attributed Network Embedding 1 2 2 1 3 Wentao Huang, Yuchen Li, Yuan Fang, Ju Fan, Hongxia Yang 1 School of Information, Renmin University of China 2 School of Information System, Singapore Management University 3 Damo


  1. BiANE: Bipartite Attributed Network Embedding 1 2 2 1 3 Wentao Huang, Yuchen Li, Yuan Fang, Ju Fan, Hongxia Yang 1 School of Information, Renmin University of China 2 School of Information System, Singapore Management University 3 Damo Academy, Alibaba Group

  2. Outline 2 q Introduction & Challenge q Methodology q Experiment q Conclusion & Future Work

  3. Introduction 3 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food 23 Barilla soccer player $8 Rio de Janeiro ⋯ ⋯ Lisa Makeup Palette female cosmetics 21 YSL actress $32 Milan ⋯ ⋯

  4. Introduction 4 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food q Characteristics 23 Barilla soccer player $8 Rio de Janeiro ⋯ ⋯ Lisa Makeup Palette female cosmetics 21 YSL actress $32 Milan ⋯ ⋯

  5. Introduction 5 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food q Characteristics 23 Barilla soccer player $8 Rio de Janeiro ⋯ The Inter-Partition Proximity § ⋯ Lisa Makeup Palette female cosmetics 21 YSL actress $32 Milan ⋯ ⋯

  6. Introduction 6 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food q Characteristics 23 Barilla soccer player $8 Rio de Janeiro ⋯ The Inter-Partition Proximity § ⋯ Lisa Makeup Palette The Intra-Partition Proximity § female cosmetics 21 YSL actress $32 Milan ⋯ ⋯

  7. Introduction 7 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food q Characteristics 23 Barilla soccer player $8 Rio de Janeiro ⋯ The Inter-Partition Proximity § ⋯ Lisa Makeup Palette The Intra-Partition Proximity § female cosmetics 21 YSL The Attribute Proximity 1) actress $32 Milan ⋯ ⋯

  8. Introduction 8 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food q Characteristics 23 Barilla soccer player $8 Rio de Janeiro ⋯ The Inter-Partition Proximity § ⋯ Lisa Makeup Palette The Intra-Partition Proximity § female cosmetics 21 YSL The Attribute Proximity 1) actress $32 Milan ⋯ The Structure Proximity ⋯ 2)

  9. Introduction 9 q Bipartite Attributed Network Mario Soccer Ball male sports utility E-Commerce Websites 18 Nike ü soccer player $48 ⋯ Rome Recommendation System ü ⋯ Alicia Cosmetic Bag Bibliometric Network Analysis ü female cosmetics 22 L‘Oreal Biological Community Detection ü model $59 ⋯ Paris ⋯ Risk Assessment of Financial Systems ü Ricardo Spaghetti male food q Characteristics 23 Barilla soccer player $8 Rio de Janeiro ⋯ The Inter-Partition Proximity § ⋯ Lisa Makeup Palette The Intra-Partition Proximity § female cosmetics 21 YSL The Attribute Proximity 1) actress $32 Milan ⋯ The Structure Proximity ⋯ 2) q Goal: Given a bipartite attributed network G =( 𝒱 , 𝒲 , E , 𝐘 𝒱 , 𝐘 𝒲 ) , we want to learn a mapping function to transform each node to a vector in a low-dimension space.

  10. Technical Challenges 10 q The Attribute-Structure Correlation § Complementarity & Coherence Mario Soccer Ball male sports utility 18 Nike soccer player $48 Rome ⋯ ⋯ Alicia Cosmetic Bag female cosmetics 22 L‘Oreal model $59 Mario ⋯ Paris ⋯ male Ricardo Spaghetti male food 18 23 Barilla soccer player $8 soccer player ⋯ Rio de Janeiro ⋯ Rome Lisa Makeup Palette female cosmetics ⋯ 21 YSL actress $32 ⋯ Milan ⋯ The Structure Information The Attribute Information q Negative Sampling Strategy § Static sampling strategies can not reflect the variation of embedding space. § Dynamic sampling strategies will result in the scalability issue.

  11. Methodology 11 ! " First-Order .3 .4 .1 .7 Structure .2 .3 .8 .5 Proximity ! " Autoencoder First-Order .9 .3 .9 .5 ! Structure Modeling Latent Dynamic ! " .1 .1 .3 .4 Proximity Autoencoder Correlation Positive ⋯ ⋯ ⋯ ⋯ Modeling Latent Dynamic Training Sampling ! First-Order Correlation Positive Attribute .3 .4 .1 .7 Proximity ! Training Sampling Autoencoder First-Order .2 .3 .8 .5 Attribute Modeling ! .9 .3 .9 .5 Proximity Autoencoder .1 .1 .3 .4 Modeling ⋯ ⋯ ⋯ ⋯ G =( ! , " , � , # ! , # " ) Intra-Partition Proximity Modeling Inter-Partition Proximity Modeling

  12. Example 12 q Scholar-Publication Network • Jiawei Han • Gender: Male • Institutions: UIUC, SFU • Research Interests: • Data Mining • Database Systems Scholar • Data Warehousing J Han P Cui WF Wong • Information Networks UIUC; ML NUS; ARCH THU; ML, AI W Zhu Y Sun WW Hwu BC Ooi T Chen D Chen THU; ML UCLA; ML UIUC; ARCH NUS; DB UCLA; ML UIUC; EDA Scholar Partition: WW Hwu: Wen-mei W. Hwu WF Wong: Weng-Fai Wong BC Ooi: Beng Chin Ooi D Chen: Deming Chen W Zhu: Wenwu Zhu T Chen: Ting Chen Y Sun: Yizhou Sun J Han: Jiawei Han P Cui: Peng Cui Publication Partition: PSL TF-IPM: Topic-Factorized Ideal Point Estimation Model for Legislative Voting Network. AccDNN ICDE IFUHJ: Is FPGA Useful for Hash Joins? FCCM TF-IPM H-tree AccDNN: An IP-Based DNN Generator for FPGAs. SIGKDD IFUHJ VLDBJ PSL: Parallelizing Skip Lists for In-Memory Multi-Core Database Systems. SDNE CIDR NetClus SDNE: Structural Deep Network Embedding. SIGKDD Publication SIGKDD H-tree: Index nesting – an efficient approach to indexing in object-oriented databases. NetClus: Ranking-Based Clustering of Heterogeneous Information Networks with Star Network Schema.

  13. Intra-Partition Proximity Modeling 13 # & % & • Jiawei Han (#) (#) ⋯ (#) ( ! " ! ( % • Gender: Male ! ! ! ⋯ ⋯ • Institutions: UIUC, SFU • Research Interests: $ # • Data Mining #′ %′ • Database Systems % $ • Data Warehousing • Information Networks (#) ⋯ (#) ⋯ (#) " % zz ! ! ⋯ ! ! % # % Scholar J Han P Cui WF Wong UIUC; ML THU; ML, AI NUS; ARCH BC Ooi Y Sun W Zhu WW Hwu BC Ooi T Chen D Chen WF Wong UCLA; ML THU; ML UIUC; ARCH NUS; DB NUS; DB UCLA; ML UIUC; EDA NUS; ARCH T Chen UCLA; ML J Han D Chen UIUC; ML UIUC; EDA PSL AccDNN ICDE FCCM TF-IPM Y Sun H-tree SIGKDD WW Hwu IFUHJ VLDBJ UCLA; ML SDNE CIDR NetClus UIUC; ARCH SIGKDD Publication SIGKDD $ ! + ⋯ ⋯+ # $ "#$ + # $ + # $ % ! = #

  14. Intra-Partition Proximity Modeling 14 & & # % (#) (#) ⋯ (#) ( ! " ! ( % ! ⋯ ! ⋯ ! $ # #′ %′ % $ (#) (#) (#) ⋯ " % ⋯ ! ! zz ! ! % ⋯ ! # % "

  15. Intra-Partition Proximity Modeling 15 & & # % (#) (#) ⋯ (#) ( ! " ! ( % ! ⋯ ! ⋯ ! $ # #′ %′ % $ (#) (#) (#) ⋯ " % ⋯ ! ! zz ! ! % ⋯ ! # % " q Compact Feature Learning

  16. Intra-Partition Proximity Modeling 16 & & # % (#) (#) ⋯ (#) ( ! " ! ( % ! ⋯ ! ⋯ ! $ # #′ %′ % $ (#) (#) (#) ⋯ " % ⋯ ! ! zz ! ! % ⋯ ! # % " q Joint Modeling — Preserving the first-order proximity

  17. Latent Correlation Training 17 & & # % (#) (#) ⋯ (#) ( ! " ! ( % ! ⋯ ! ⋯ ! $ # #′ %′ % $ (#) (#) (#) ⋯ " % ⋯ ! ! zz ! ! % ⋯ ! # % "

  18. Latent Correlation Training 18 & & # % (#) (#) ⋯ (#) ( ! " ! ( % ! ⋯ ! ⋯ ! $ # #′ %′ % $ (#) (#) (#) ⋯ " % ⋯ ! ! zz ! ! % ⋯ ! # % " q Transform encodings to latent representations via auxiliary kernels.

Recommend


More recommend