Graph Embedding and Reasoning

Jie Tang Department of Computer Science and Technology Tsinghua University

The slides can be downloaded at http://keg.cs.tsinghua.edu.cn/jietang 1 Networked World

• 2 billion MAU • >777 million trans. (alipay) • 26.4 billion minutes/day • 200 billion on 11/11

• 320 million MAU • 350 million users • Peak: 143K tweets/s • influencing our daily life

• 700 million MAU • ~200 million MAU • 95 million pics/day • 70 minutes/user/day

• 300 million MAU •QQ: 860 million MAU • 30 minutes/user/day • WeChat: 1 billion MAU

2 Mining Big Graphs/Networks

• An information/social graph is made up of a set of individuals/entities (“nodes”) tied by one or more interdependency (“edges”), such as friendship.

Mining big networks: “A field is emerging that leverages the capacity to collect and analyze data at a scale that may reveal patterns of individual and group behaviors.” [1]

1. David Lazer, Alex Pentland, Lada Adamic, Sinan Aral, Alber-Laszlo Barabasi, et al. Computational Social Science. Science 2009. 3 Challenges

Info. Space + Social Space Challenges

Info. big Space

dynamic Interaction

Social hetero Space geneous Interaction

1. J. Scott. (1991, 2000, 2012). Social network analysis: A handbook. 2. D. Easley and J. Kleinberg. Networks, crowds, and markets: Reasoning about a highly connected world. Cambridge University Press, 2010. 4 Social & Information Network Analysis

o Graph Neural Networks o Deep Learning for Networks Recent Trend: o High-Order Networks [Benson et al.] 2015~2019 Deep Learning for Networks o Info. vs. Social Networks (Twitter) [Kwak et al.] o Signed Networks [Leskovec et al.] o Semantic Social Networks [Tang et al.] 2010~2014 o Graph Evolution [Leskovec et al.] o Four Deg. Of Separation [Backstrom et al.] o 3 Deg. Of Influence [Christakis & Fowler] The 1st decade of the 21st o Structural Diversity [Ugander et al.] o Social Influence Analysis [Tang et al.] o Computational Social Science [Watts] 2005~2009 o Six Deg. Of Separation [Leskovec & Horvitz] Century: o Network Embedding [Perozzi et al.] o Network Heterogeneity [Sun & Han] More Computer & Data o Network Embedding [Tang & Liu] Scientists o Influence Max’n [Domingos & Kempe et al.] 2000~2004 o Computer Social Science [Lazer et al.] o Community Detection [Girvan & Newman] o Network Motifs [Milo et al.] 1998/9 o Small Worlds [Watts & Strogatz] o Link Prediction [Liben-Nowell & Kleinberg] o Scale Free [Barabasi & Albert] The Late 20th Century: 1997 o Power Law [Faloutsos 3] o HITS [Kleinberg] o PageRank [Page & Brin] 1992 o Structural Hole [Burt] CS & Physics o Hyperlink Vector Voting [Li] o Dunbar’s Number [Dunbar] 1970s o The Strength Of Weak Tie [Granovetter] o Small Worlds [Migram] 1960s th 1950s o Homophily [Lazarsfeld & Merton] The 20 Century: o Random Graph [Erdos, Renyi, Gilbert] o Balance Theory [Heider et al.] o Degree Sequence [Tuttle, Havel, Hakami] Sociology & Anthropology 1930s o Sociogram [Moreno]

5 Recent Trend

d-dimensional vector, d<<|V| Representation Learning/ Graph Embedding 0.8 0.2 0.3 … 0.0 0.0

Users with the same label are located in the d-dimensional space closer than those with different labels

e.g., node classification

label2

label1

6 Example1: Open Academic Graph —how can we benefit from graph embedding?

Open Academic Graph (OAG) is a large knowledge graph by linking two billion-scale academic graphs: Microsoft Academic Graph (MAG) and AMiner.

240M Authors 220M Publications Event • 189M publications Open • 113M researchers Academic • 8M concepts Graph • 8M IP accesses @ Open data knowledge intelligence Academic 25K Institutions Society 664K Fields of Studies 50K Venues Tsinghua University AMiner Microsoft Academic Graph

Linking large-scale heterogeneous academic graphs https://www.openacademic.ai/oag/

1. F. Zhang, X. Liu, J. Tang, Y. Dong, P. Yao, J. Zhang, X. Gu, Y. Wang, B. Shao, R. Li and K. Wang. OAG: Toward Linking Large-scale Heterogeneous Entity Graphs. KDD'19. 7 Results

Our methods based on graph embeddings

** Code&Data available at https://github.com/zfjsail/OAG 8 Example2: Social Prediction —how can we benefit from graph embedding?

Example: Online Forum

E E D D F F H v H v1 v2 B B C A C A

Who are more likely to be “active”, v1 or v2? e.g., active = “dropout an online forum”

Active neighbor Inactive neighbor v User to be influenced

1. J. Qiu, J. Tang, H. Ma, Y. Dong, K. Wang, and J. Tang. DeepInf: Social Influence Prediction with Deep Learning. KDD'18. 9 Dropout Prediction

• Dropout prediction with influence – Problem: dropout prediction – Game data: King of Glory (王者荣耀)

With Influence w/o influence Top k #message #success ratio #message #success ratio 1 3996761 1953709 48.88% 6617662 2732167 41.29% 2 2567279 1272037 49.55% 9756330 4116895 42.20% 3 1449256 727728 50.21% 10537994 4474236 42.46% 4 767239 389588 50.78% 9891868 4255347 43.02% 5 3997251 2024859 50.66% 15695743 6589022 41.98%

10 Outline

• Representation Learning for Graphs – Unified graph embedding as matrix factorization – Scalable graph embedding – Fast graph embedding • Extensions and Reasoning – Dynamic – Heterogeneous – Reasoning • Conclusion and Q&A

11 Representation Learning on Networks

푺 = 푓(푨) RLN by Matrix

achieve better accuracy Input: Output: Adjacency Matrix Sparsify 푺 Scalable RLN Vectors 푨 풁 handle 100M graph

Fast RLN 풁 = 푓(풁′)

offer 10-400X speedups

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 12 DeepWalk

SkipGram with Random walk One example RW path Hierarchical softmax

v3 v4 v4 v3 v1 v5 v6

v 1 v5

v2 v6

Hierarchical softmax

1. B. Perozzi, R. Al-Rfou, and S. Skiena. 2014. Deepwalk: Online learning of social representations. KDD, 701–710. 13 Later…

• LINE[1]: explicitly preserves both first- order and second-order proximities.

• PTE[2]: learn heterogeneous text network embedding via a semi- supervised manner.

• Node2vec[3]: use a biased random walk to better explore node’s neighborhood.

1. J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan, and Q. Mei. 2015. Line: Large-scale information network embedding. WWW, 1067–1077. 2. J. Tang, M. Qu, and Q. Mei. 2015. Pte: Predictive text embedding through large-scale heterogeneous text networks. KDD, 1165–1174. 3. A. Grover and J. Leskovec. 2016. node2vec: Scalable feature learning for networks. KDD, 855–864. 14 Questions

• What are the fundamentals underlying the different models? or • Can we unify the different network embedding approaches?

15 Unifying DeepWalk, LINE, PTE, and node2vec into Matrix Forms

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 16 Starting with DeepWalk

17 DeepWalk Algorithm

18 Skip-gram with Negative Sampling

• SGNS maintains a multiset 퓓 that counts the occurrence of each word-context pair (푤, 푐) • Objective 푏# 푤 #(푐) ℒ = ෍ ෍(# 푤, 푐 log 푔 푥푇 푥 + log 푔(−푥푇 푥 )) 푤 푐 |풟| 푤 푐 푤 푐

where xw and xc are d-dimenational vector • For sufficiently large dimension d, the objective above is equivalent to factorizing the PMI matrix[1] #(푤, 푐)|풟| log 푏#(푤)#(푐)

1. Levy and Goldberg. Neural word embeddings as implicit matrix factorization. In NIPS 2014 19 DeepWalk

20 Understanding random walk + skip gram

Suppose the multiset 풟 is constructed based on random #(푤,푐)|풟| walk on graphs, can we interpret 푙표푔 with graph 푏#(푤)#(푐) structures?

21 Understanding random walk + skip gram

• Partition the multiset 풟 into several sub-multisets according to the way in which each node and its context appear in a random walk node sequence. • More formally, for 푟 = 1, 2, ⋯ , 푇, we define

Distinguish direction and distance

22 Understanding random walk + skip gram

23 Understanding random walk + skip gram

24 Understanding random walk + skip gram

the length of random walk 퐿 → ∞

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 25 Understanding random walk + skip gram

the length of random walk 퐿 → ∞

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 26 Understanding random walk + skip gram

the length of random walk 퐿 → ∞

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 27 Understanding random walk + skip gram

the length of random walk 퐿 → ∞

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 28 29 DeepWalk is factorizing…

푤푖−2

푤푖−1 푤푖

푤푖+1

푤푖+2

DeepWalk is asymptotically and implicitly factorizing

푣표푙 퐺 = ෍ ෍ 퐴푖푗 푖 푗

푨 Adjacency matrix b: #negative samples 푫 Degree matrix T: context window size

30 LINE

31 PTE

32 node2vec — 2nd Order Random Walk

33 Unifying DeepWalk, LINE, PTE, and node2vec into Matrix Forms

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 34 • Can we directly factorize the derived matrix?

35 NetMF: explicitly factorizing the DW matrix

푤푖−2

Matrix 푤푖−1 푤푖 Factorization푤푖+1 푤푖+2

DeepWalk is asymptotically and implicitly factorizing

1. Qiu et al. Network embedding as matrix factorization: unifying deepwalk, line, pte, and node2vec. WSDM’18. The most cited paper in WSDM’18 as of May 2019 36 Explicitly factorize the matrix

* approximate D-1/2AD-1/2 with T its top-h eigenpairs Uh흠hUh

* decompose using Arnoldi algorithm[1]

1. R. Lehoucq, D. Sorensen, and C.Yang. 1998. ARPACK users’ guide: solution of large-scale eigenvalue problems with implicitly restarted Arnoldi methods. SIAM. 37 Error Bound Analysis for NetMF

• According to Frobenius norm’s property

• and because M’i,j=max(Mi,j, 1)>=1, we have

• Also because the property of NGL,

38 Experimental Setup

** Code available at https://github.com/xptree/NetMF 39 Results

40 Results

• NetMF significantly outperforms LINE and DeepWalk when sparse labeled vertices are provided

41 Challenge in NetMF

Small world Academic graph

42 Representation Learning on Networks

푺 = 푓(푨) RLN by Matrix

achieve better accuracy Input: Output: Adjacency Matrix Sparsify 푺 Scalable RLN Vectors 푨 풁 handle 100M graph

Fast RLN 풁 = 푓(풁′)

offer 10-400X speedups

1. J. Qiu, Y. Dong, H. Ma, J. Li, C. Wang, K. Wang, and J. Tang. NetSMF: Large-Scale Network Embedding as Sparse Matrix Factorization. WWW’19. 43 Sparsify 푺

For random-walk matrix polynomial

where and non-negative

One can construct a 1 + 휖 -spectral sparsifier 푳෨ with non-zeros

in time for undirected graphs

1. D. Cheng, Y. Cheng, Y. Liu, R. Peng, and S.H. Teng, Efficient Sampling for Gaussian Graphical Models via Spectral Sparsification, COLT 2015. 2. D. Cheng, Y. Cheng, Y. Liu, R. Peng, and S.H. Teng. Spectral sparsification of random-walk matrix polynomials. arXiv:1502.03496. 44 Sparsify 푺

For random-walk matrix polynomial

where and non-negative

One can construct a 1 + 휖 -spectral sparsifier 푳෨ with non-zeros

in time

45 Sparsify 푺

For random-walk matrix polynomial

where and non-negative

One can construct a 1 + 휖 -spectral sparsifier 푳෨ with non-zeros

in time

46 Sparsify 푺

For random-walk matrix polynomial

where and non-negative

One can construct a 1 + 휖 -spectral sparsifier 푳෨ with non-zeros

in time

47 NetSMF --- Sparse

Factorize the constructed matrix

1. J. Qiu, Y. Dong, H. Ma, J. Li, C. Wang, K. Wang, and J. Tang. NetSMF: Large-Scale Network Embedding as Sparse Matrix Factorization. WWW’19. 48 NetSMF — Approximation Error

49 Datasets

** Code available at https://github.com/xptree/NetSMF 50 Results

** Code available at https://github.com/xptree/NetSMF 51 Representation Learning on Networks

푺 = 푓(푨) RLN by Matrix

achieve better accuracy Input: Output: Adjacency Matrix Sparsify 푺 Scalable RLN Vectors 푨 풁 handle 100M graph

Fast RLN 풁 = 푓(풁′)

offer 10-400X speedups

1. J. Zhang, Y. Dong, Y. Wang, J. Tang, and M. Ding. ProNE: Fast and Scalable Network Representation Learning. IJCAI’19. 52 ProNE: Fast and Scalable Network Embedding

1. J. Zhang, Y. Dong, Y. Wang, J. Tang, and M. Ding. ProNE: Fast and Scalable Network Representation Learning. IJCAI’19. 53 ProNE: Fast and Scalable Network Embedding

* ProNE (1 thread) v.s. Others (20 threads)

* 10 minutes on Youtube (~1M nodes)

** Code available at https://github.com/THUDM/ProNE 54 NE as Sparse Matrix Factorization

• node-context set (sparsity)

• For edge (i, j), the occurrence probability of context j given node i

• objective (weighted sum of log-loss)

55 NE as Sparse Matrix Factorization

• To avoid the trivial solution

• Local negative samples drawn from

• Modify the loss (sum over the edge-->sparse)

56 NE as Sparse Matrix Factorization

• Let the partial derivative w.r.t. be zero

• Matrix to be factorized (sparse)

57 NE as Sparse Matrix Factorization

• Matrix to be factorized (sparse)

• Many methods for truncated SVD with the sparse matrix (e.g., randomized tSVD, )

• This optimization method (single thread) is much faster than SGD used in DeepWalk, LINE, etc. and is still scalable!!!

• Orthogonality, avoid direct non-convex optimization

58 NE as Sparse Matrix Factorization

• Compared with matrix factorization method (e.g., NetMF)

V.S.

• Sparsity (local structure and local negative samples)→ much faster and scalable

• Challenge: may lose high order information!

• Improvement via spectral propagation

59 Higher-order Cheeger’s inequality

•

• Bridge graph spectrum and graph partitioning

• k-way Cheeger constant :reflects the effect of the graph partitioned into k parts. A smaller value of means a better partitioning effect.

1. J. R. Lee, S. O. Gharan, L. Trevisan. Multiway spectral partitioning and higher-order cheeger inequalities. Journal of the ACM (JACM), 2014, 61(6): 37. 60 Higher-order Cheeger’s inequality

• Bridge graph spectrum and graph partition

• low eigenvalues control the global information

• high eigenvalues control the local information

•

• spectral propagation: propagate the initial network embedding in the spectrally modulated network !

61 NE Enhancement via Spectral Propagation

• signal filter low pass filter band pass filter high pass filter

62 NE Enhancement via Spectral Propagation

• signal filter (low-pass example)

63 NE Enhancement via Spectral Propagation

• spectral propagation for NE

• is the spectral filter of

• is modulated by the filter in the spectrum

64 NE Enhancement via Spectral Propagation

• the form of the spectral filter

• Band-pass (low-pass, high-pass)

• pass eigenvalues within a certain range and weaken eigenvalues outside that range

• amplify local and global network information

65 Chebyshev Expansion for Efﬁciency

• Utilize truncated Chebyshev expansion to avoid explicit eigendecomposition and Fourier transform

• Calculate the coefﬁcient of Chebyshev expansion

where is the modified Bessel function of the first kind 66 Chebyshev Expansion for Efficiency

• NE

• Denote and calculate the Equation in the following recursive way

67 Complexity of ProNE

• Spectral propagation only involves sparse matrix multiplication! The complexity is linear!

• sparse matrix factorization + spectral propagation = O(|V|d2 + k|E|)

68 Datasets

** Code available at https://github.com/THUDM/ProNE 69 Results

* ProNE (1 thread) v.s. Others (20 threads)

* 10 minutes on Youtube (~1M nodes)

** Code available at https://github.com/THUDM/ProNE 70 Effectiveness experiments

* ProNE (SMF) = ProNE w/ only sparse matrix factorization Embed 100,000,000 nodes by one thread: 29 hours with performance superiority

** Code available at https://github.com/THUDM/ProNE 71 Spectral Propagation: k-way Cheeger

72 Outline

73 BurstGraph: Dynamic Network Embedding with Burst Detection

• GraphSage aggregates structural information for each snapshot of dynamic networks. • VAE reconstructs the vanilla evolution and bursty evolution. • RNN Structure extends the model for dynamic networks.

1. Y. Zhao, X. Wang, H. Yang, L. Song, and J. Tang. Large Scale Evolving Graphs with Burst Detection. IJCAI’19. 74 Information Aggregation

• To get the feature information of each snapshot, we utilize GraphSage to aggregate the hidden representation from vertices’ neighbors.

• In each network snapshot, GraphSage samples neighbors via uniform distribution.

• Then GraphSage uses simple max-pooling operator to aggregate information from neighbors.

75 Vanilla Evolution

• Based on the feature information 퐺푖,푡 aggregated by GraphSAGE, vanilla and bursty evolution both choose VAE to reconstruct the structure of network snapshot;

• The random variable 풛푡 of vanilla evolution follows a normal Gaussian distribution:

76 Bursty Evolution

• Due to the sparsity and contingency of bursty evolution, the random variable 풔푡 of bursty evolution follows a Spike-and-Slab distribution:

77 Temporal Consistence

• Taking the temporal structure of the dynamic networks into account, we introduce an RNN structure into our framework.

78 Link Prediction Result

• Data – Alibaba: one of the largest E-commerce platform • Methods – Static NE: DeepWalk, GraphSage – Dynamic NE: TNE, CTDNE

* Task: predicting links between users and items

** Code available at https://github.com/ericZhao93/BurstGraph 79 Outline

80 Multiplex Heterogeneous Graph Embedding

• Multiplex networks comprise not only multi-typed nodes and/or edges but also a rich set of attributes.

• How do we deal with the multiplex edges?

click add-to-preference add-to-cart Gender: male Age: 23 conversion Price: $1000 Location: Beijing Brand: Lenovo … …

Gender: female Age: 26 Price: $800 Location: Bangalore Brand:Apple … …

Gender: male Age: 35 Price: $50 Location: Boston Brand: Nike … … users items

1. Y. Cen, X. Zou, J. Zhang, H. Yang, J. Zhou and J. Tang. Representation Learning for Attributed Multiplex Heterogeneous Network. KDD’19. 81 Related Network Embedding Methods

82 GATNE: Multiplex Heterogeneous Graph Embedding

Base Embedding

Output Layer GATNE-T GATNE-I Input Layer … 0 Hidden 0 Layer 0 Node Attributes 0 |V1|×K1 0 1 0 0 1 0 0 0 1 0 0 1 1 … 0 0 1 0 1 0 0 0 |V2|×K2 0 0 0 Edge Embedding … |V|-dim

|V3|×K3

Generating Overall Node Embedding Heterogeneous Skip-Gram

1. Y. Cen, X. Zou, J. Zhang, H. Yang, J. Zhou and J. Tang. Representation Learning for Attributed Multiplex Heterogeneous Network. KDD’19. 83 Link Prediction Results

• Data – Small: Amazon, YouTube, Twitter w/ 10K nodes – Large: Alibaba w/ 40M nodes and 0.5B edges

GATNE outperforms all sorts of baselines in the various datasets.

** Code available at https://github.com/THUDM/GATNE 84 Outline

85 Cognitive Graph: DL + Dual Process Theory

explicit decision

For more, please come to the DS in China Session tomorrow morning (Summit Room 5/6)

implicit knowledge expansion

1. M. Ding, C. Zhou, Q. Chen, H. Yang, and J. Tang. Cognitive Graph for Multi-Hop Reading Comprehension at Scale. ACL’19. 86 CogDL —A Toolkit for Deep Learning on Graphs

** Code available at https://keg.cs.tsinghua.edu.cn/cogdl/ 87 CogDL —A Toolkit for Deep Learning on Graphs

88 CogDL features

• CogDL supports these features: ✓ Sparsification: fast network embedding on large-scale networks with tens of millions of nodes ✓ Arbitrary: dealing with different graph structures: attributed, multiplex and heterogeneous networks ✓ Distributed: multi-GPU training on one machine or across multiple machines ✓ Extensible: easily register new datasets, models, criterions and tasks

• Homepage: https://keg.cs.tsinghua.edu.cn/cogdl/

89 CogDL Leaderboards

Multi-label Node Classification The following leaderboard is built from unsupervised learning with multi-label node classification setting.

http://keg.cs.tsinghua.edu.cn/cogdl/node-classification.html 91 CogDL Leaderboards

Node Classification with Attributes.

http://keg.cs.tsinghua.edu.cn/cogdl/node-classification.html 92 Leaderboards: Link Prediction

http://keg.cs.tsinghua.edu.cn/cogdl/link-prediction.html 93 Join us

• Feel free to join us with the three following ways: ✓ add your data into the leaderboard ✓ add your algorithm into the leaderboard ✓ add your algorithm into the toolkit

• Note that, in order to collaborate with us, you may refer to the documentation in our homepage.

94 Graph Embedding: A Quick Summary

Cognitive Graph 2019: Ding et al., ACL’19 ProNE, GATNE, BurstGraph 2019: Zhang et al., IJCAI’19; Cen et al., KDD’19; Zhao et al., IJCAI’19

FastGCNs, Graph attention 2018: Velickovic et al., ICLR’18, Chen et al., ICLR 2018 NetMF & NetSMF 2018: Qiu et al., WSDM’18 & WWW’19

Neural message passing, GraphSage 2017: Gilmer et al., ICML’17; Hamilton et al., NIPS’17

Gated graph neural network 2016: Li et al., ICLR’16 structure2vec 2016: Dai et al., ICML’16

Graph convolutional network 2015: Duvenaud et al., NIPS’15; Kipf & Welling ICLR’17 PTE, metapath2vec 2015: Tang et al., KDD’15; Dong et al., KDD’17 LINE, node2vec 2015: Tang et al., WWW’15; Grover & Leskovec, KDD’16

DeepWalk 2014: Perozzi et al., KDD’14 Spectral graph convolution 2014: Bruna et al., ICLR’14

word2vec (skip-gram) 2013: Mikolov et al., ICLR’13

Graph neural network 2005: Gori et al., IJCNN’05

Spectral clustering 2000: Ng et al. & Shi, Malik

Spectral partitioning 1973: Donath & Hoffman

95 Representation Learning on Networks

CogDL: A Toolkit for Deep Learning on Graphs

푺 = 푓(푨) RLN by Matrix

achieve better accuracy Input: Output: Adjacency Matrix Sparsify 푺 Scalable RLN Vectors 푨 풁 handle 100M graph

Fast RLN 풁 = 푓(풁′)

offer 10-400X speedups

Dynamic & Cognitive Graph Heterogeneous

96 Related Publications

• Jie Zhang, Yuxiao Dong, Yan Wang, Jie Tang, and Ming Ding. ProNE: Fast and Scalable Network Representation Learning. IJCAI’19. • Yukuo Cen, Xu Zou, Jianwei Zhang, Hongxia Yang, Jingren Zhou and Jie Tang. Representation Learning for Attributed Multiplex Heterogeneous Network. KDD’19. • Fanjin Zhang, Xiao Liu, Jie Tang, Yuxiao Dong, Peiran Yao, Jie Zhang, Xiaotao Gu, Yan Wang, Bin Shao, Rui Li, and Kuansan Wang. OAG: Toward Linking Large-scale Heterogeneous Entity Graphs. KDD’19. • Yifeng Zhao, Xiangwei Wang, Hongxia Yang, Le Song, and Jie Tang. Large Scale Evolving Graphs with Burst Detection. IJCAI’19. • Yu Han, Jie Tang, and Qian Chen. Network Embedding under Partial Monitoring for Evolving Networks. IJCAI’19. • Ming Ding, Chang Zhou, Qibin Chen, Hongxia Yang, and Jie Tang. Cognitive Graph for Multi-Hop Reading Comprehension at Scale. ACL’19. • Jiezhong Qiu, Yuxiao Dong, Hao Ma, Jian Li, Chi Wang, Kuansan Wang, and Jie Tang. NetSMF: Large-Scale Network Embedding as Sparse Matrix Factorization. WWW'19. • Jiezhong Qiu, Jian Tang, Hao Ma, Yuxiao Dong, Kuansan Wang, and Jie Tang. DeepInf: Modeling Influence Locality in Large Social Networks. KDD’18. • Jiezhong Qiu, Yuxiao Dong, Hao Ma, Jian Li, Kuansan Wang, and Jie Tang. Network Embedding as Matrix Factorization: Unifying DeepWalk, LINE, PTE, and node2vec. WSDM’18. • Jie Tang, Jing Zhang, Limin Yao, Juanzi Li, Li Zhang, and Zhong Su. ArnetMiner: Extraction and Mining of Academic Social Networks. KDD’08.

For more, please check here http://keg.cs.tsinghua.edu.cn/jietang 97 Thank you！ Collaborators: Jie Zhang, Jiezhong Qiu, Ming Ding, Qibin Chen, Yifeng Zhao, Yukuo Cen, Yu Han, Fanjin Zhang, Xu Zou, Yan Wang, et al. (THU) Yuxiao Dong, Chi Wang, Hao Ma, Kuansan Wang (Microsoft) Hongxia Yang, Chang Zhou, Le Song, Jingren Zhou, et al. (Alibaba)

Jie Tang, KEG, Tsinghua U http://keg.cs.tsinghua.edu.cn/jietang Download all data & Codes https://keg.cs.tsinghua.edu.cn/cogdl/