SlideShare une entreprise Scribd logo
1  sur  71
Télécharger pour lire hors ligne
SSumM : Sparse Summarization
of Massive Graphs
Kyuhan Lee* Hyeonsoo Jo* Jihoon Ko Sungsu Lim Kijung Shin
Graphs are Everywhere
Citation networksSubway networks Internet topologies
Introduction Algorithms Experiments ConclusionProblem
Designed by Freepik from Flaticon
Massive Graphs Appeared
Social networks
2.49 Billion active users
Purchase histories
0.5 Billion products
World Wide Web
5.49 Billion web pages
Introduction Algorithms Experiments ConclusionProblem
Difficulties in Analyzing Massive graphs
Computational cost
(number of nodes & edges)
Introduction Algorithms Experiments ConclusionProblem
Difficulties in Analyzing Massive graphs
>
Cannot fit
Introduction Algorithms Experiments ConclusionProblem
Input Graph
Solution: Graph Summarization
>
Introduction Algorithms Experiments ConclusionProblem
Can fit
Summary Graph
Advantages of Graph Summarization
Introduction Algorithms Experiments ConclusionProblem
• Many graph compression techniques are available
• TheWebGraph Framework [BV04]
• BFS encoding [AD09]
• SlashBurn [KF11]
• VoG [KKVF14]
• Graph summarization stands out because
• Elastic: reduce size of outputs as much as we want
• Analyzable: existing graph analysis and tools can be applied
• Combinable for Additional Compression: can be further compressed
Example of Graph Summarization
1
2
3 4
5
6
7
8
9
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Adjacency Matrix
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Introduction Algorithms Experiments ConclusionProblem
Input Graph
Example of Graph Summarization
1
2
3 4
5
6
7
8
9
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Adjacency Matrix
Input Graph
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
𝑽𝑽𝟏𝟏={1,2}
𝑽𝑽𝟕𝟕={7,8,9}
𝑽𝑽𝟑𝟑={3,4,5,6}
Summary Graph
Introduction Algorithms Experiments ConclusionProblem
Subnode
Subedge
Supernode
Example of Graph Summarization
1
2
3 4
5
6
7
8
9
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Adjacency Matrix
Input Graph
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
𝑽𝑽𝟏𝟏={1,2}
𝑽𝑽𝟕𝟕={7,8,9}
𝑽𝑽𝟑𝟑={3,4,5,6}
Summary Graph
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3
Introduction Algorithms Experiments ConclusionProblem
Subnode
Subedge
Superedge
Supernode
Example of Graph Summarization
1
2
3 4
5
6
7
8
9
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Adjacency Matrix
Input Graph
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
𝑽𝑽𝟏𝟏={1,2}
𝑽𝑽𝟕𝟕={7,8,9}
𝑽𝑽𝟑𝟑={3,4,5,6}
Summary Graph
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3
𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟕𝟕} =2
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1
𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3
𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5
Introduction Algorithms Experiments ConclusionProblem
Example of Graph Summarization
1
2
3 4
5
6
7
8
9
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Adjacency Matrix
Input Graph
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
𝑽𝑽𝟏𝟏={1,2}
𝑽𝑽𝟕𝟕={7,8,9}
𝑽𝑽𝟑𝟑={3,4,5,6}
Summary Graph
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1
𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3
𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5
Introduction Algorithms Experiments ConclusionProblem
Example of Graph Summarization
Introduction Algorithms Experiments ConclusionProblem
1 2 3 4 5 6 7 8 9
1 0 1 3/8 3/8 3/8 3/8 1/3 1/3 1/3
2 1 0 3/8 3/8 3/8 3/8 1/3 1/3 1/3
3 3/8 3/8 0 5/6 5/6 5/6 0 0 0
4 3/8 3/8 5/6 0 5/6 5/6 0 0 0
5 3/8 3/8 5/6 5/6 0 5/6 0 0 0
6 3/8 3/8 5/6 5/6 5/6 0 0 0 0
7 1/3 1/3 0 0 0 0 0 1 1
8 1/3 1/3 0 0 0 0 1 0 1
9 1/3 1/3 0 0 0 0 1 1 0
Reconstructed Adjacency Matrix
𝑽𝑽𝟏𝟏={1,2}
𝑽𝑽𝟕𝟕={7,8,9}
𝑽𝑽𝟑𝟑={3,4,5,6}
Summary Graph
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1
𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3
𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5
Example of Graph Summarization
𝑽𝑽𝟏𝟏={1,2}
𝑽𝑽𝟕𝟕={7,8,9}
𝑽𝑽𝟑𝟑={3,4,5,6}
Summary Graph
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1
𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3
𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5
Introduction Algorithms Experiments ConclusionProblem
1 2 3 4 5 6 7 8 9
1 0 1 3/8 3/8 3/8 3/8 1/3 1/3 1/3
2 1 0 3/8 3/8 3/8 3/8 1/3 1/3 1/3
3 3/8 3/8 0 5/6 5/6 5/6 0 0 0
4 3/8 3/8 5/6 0 5/6 5/6 0 0 0
5 3/8 3/8 5/6 5/6 0 5/6 0 0 0
6 3/8 3/8 5/6 5/6 5/6 0 0 0 0
7 1/3 1/3 0 0 0 0 0 1 1
8 1/3 1/3 0 0 0 0 1 0 1
9 1/3 1/3 0 0 0 0 1 1 0
Reconstructed Adjacency Matrix
𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} = 3
𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝑜𝑜𝑜𝑜 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 8
Road Map
• Introduction
• Problem <<
• Proposed Algorithm: SSumM
• Experimental Results
• Conclusions
Problem Definition: Graph Summarization
𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮:
a graph 𝑮𝑮 and the target number of node 𝑲𝑲
𝑭𝑭𝑭𝑭𝑭𝑭𝑭𝑭:
a summary graph �𝑮𝑮
𝑻𝑻𝑻𝑻 𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴:
the difference between graph 𝑮𝑮and the restored graph �𝑮𝑮
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒕𝒕𝒕𝒕:
the number of supernodes in �𝑮𝑮 ≤ 𝑲𝑲
Introduction Algorithms Experiments ConclusionProblem
Problem Definition: Graph Summarization
𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮:
a graph 𝑮𝑮 and the target number of node 𝑲𝑲
𝑭𝑭𝑭𝑭𝑭𝑭𝑭𝑭:
a summary graph �𝑮𝑮
𝑻𝑻𝑻𝑻 𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴:
the difference between graph 𝑮𝑮and the restored graph �𝑮𝑮
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒕𝒕𝒕𝒕:
the number of supernodes in �𝑮𝑮 ≤ 𝑲𝑲
Shouldn’t we
consider sizes?
Introduction Algorithms Experiments ConclusionProblem
Problem Definition: Graph Summarization
𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮:
a graph 𝑮𝑮 and the desired size 𝑲𝑲 (in bits)
𝑭𝑭𝑭𝑭𝑭𝑭𝑭𝑭:
a summary graph �𝑮𝑮
𝑻𝑻𝑻𝑻 𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴:
the difference with graph graph 𝑮𝑮and the restored graph �𝑮𝑮
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒕𝒕𝒕𝒕:
size of �𝑮𝑮 in bits ≤ 𝑲𝑲
Introduction Algorithms Experiments ConclusionProblem
Details: Size in Bits of a Graph
𝐸𝐸 : set of edges
𝑉𝑉 : set of nodes
Encoded using log2|𝑉𝑉| bits
Introduction Algorithms Experiments ConclusionProblem
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 2 𝐸𝐸 log2 𝑉𝑉
Input graph 𝑮𝑮
Details: Size in Bits of a Summary Graph
4
5
1
4
1
5
Introduction Algorithms Experiments ConclusionProblem
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆
𝑆𝑆 : set of supernodes
𝑃𝑃 : set of superedges
𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight
Summary graph �𝑮𝑮
Details: Size in Bits of a Summary Graph
4
5
1
4
1
5
Introduction Algorithms Experiments ConclusionProblem
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆
𝑆𝑆 : set of supernodes
𝑃𝑃 : set of superedges
𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight
Summary graph �𝑮𝑮
Details: Size in Bits of a Summary Graph
4
5
1
4
1
5
Introduction Algorithms Experiments ConclusionProblem
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆
𝑆𝑆 : set of supernodes
𝑃𝑃 : set of superedges
𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight
Summary graph �𝑮𝑮
Details: Size in Bits of a Summary Graph
4
5
1
4
1
5
Introduction Algorithms Experiments ConclusionProblem
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆
𝑆𝑆 : set of supernodes
𝑃𝑃 : set of superedges
𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight
Summary graph �𝑮𝑮
Details: Error Measurement
Introduction Algorithms Experiments ConclusionProblem
1 2 3 4 5 6 7 8 9
1 0 1 3/8 3/8 3/8 3/8 1/3 1/3 1/3
2 1 0 3/8 3/8 3/8 3/8 1/3 1/3 1/3
3 3/8 3/8 0 5/6 5/6 5/6 0 0 0
4 3/8 3/8 5/6 0 5/6 5/6 0 0 0
5 3/8 3/8 5/6 5/6 0 5/6 0 0 0
6 3/8 3/8 5/6 5/6 5/6 0 0 0 0
7 1/3 1/3 0 0 0 0 0 1 1
8 1/3 1/3 0 0 0 0 1 0 1
9 1/3 1/3 0 0 0 0 1 1 0
1 2 3 4 5 6 7 8 9
1 0 1 0 0 0 1 0 0 1
2 1 0 1 0 0 1 1 0 0
3 0 1 0 1 1 1 0 0 1
4 0 0 1 0 1 0 0 0 0
5 0 0 1 1 0 1 1 0 0
6 1 1 1 0 1 0 0 0 0
7 0 1 0 0 1 0 0 1 1
8 0 0 0 0 0 0 1 0 1
9 1 0 1 0 0 0 1 1 0
Reconstructed Adjacency Matrix �𝑨𝑨Reconstructed Adjacency Matrix 𝑨𝑨
𝑅𝑅𝐸𝐸𝑝𝑝(𝑨𝑨, �𝑨𝑨) = �
𝑖𝑖=1
𝑉𝑉
�
𝑗𝑗=1
𝑉𝑉
𝐴𝐴 𝑖𝑖, 𝑗𝑗 − ̂𝐴𝐴 𝑖𝑖, 𝑗𝑗
𝑝𝑝
𝟏𝟏
𝒑𝒑
Road Map
• Introduction
• Problem
• Proposed Algorithm: SSumM <<
• Experimental Results
• Conclusions
Main ideas of SSumM
Introduction Algorithms Experiments ConclusionProblem
Combines node grouping and edge sparsification
Prunes search space
Balances error and size of the summary graph using MDL principle
• Practical graph summarization problem
◦ Given: a graph 𝑮𝑮
◦ Find: a summary graph �𝑮𝑮
◦ To minimize: the difference between 𝑮𝑮and the restored graph �𝑮𝑮
◦ Subject to: 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑜𝑜𝑜𝑜 �𝑮𝑮 in bits ≤ 𝑲𝑲
Main Idea: Combining Two Strategies
Introduction Algorithms Experiments ConclusionProblem
Node Grouping Sparsification
Main Idea: Combining Two Strategies
Introduction Algorithms Experiments ConclusionProblem
Node Grouping Sparsification
Main Idea: Combining Two Strategies
Introduction Algorithms Experiments ConclusionProblem
Node Grouping Sparsification
Main Idea: Combining Two Strategies
Introduction Algorithms Experiments ConclusionProblem
Node Grouping Sparsification
Main Idea: MDL Principle
Introduction Algorithms Experiments ConclusionProblem
1
2
3
4
5
6
Merge (5, 6)
1
2
3
4
Merge (1, 2)
3
4
5
6
5
6
1
2
{1,2}
{5,6}Merge (1, {5,6})
Merge (1, 2)
Merge (1, 3)
Merge (1, 3)
How to choose a next action?
Main Idea: MDL Principle
Introduction Algorithms Experiments ConclusionProblem
1
2
3
4
5
6
Merge (5, 6)
1
2
3
4
Merge (1, 2)
3
4
5
6
5
6
1
2
{1,2}
{5,6}Merge (1, {5,6})
Merge (1, 2)
Merge (1, 3)
Merge (1, 3)
Graph Summarization is
A Search Problem
How to choose a next action?
Main Idea: MDL Principle
Introduction Algorithms Experiments ConclusionProblem
1
2
3
4
5
6
Merge (5, 6)
1
2
3
4
Merge (1, 2)
3
4
5
6
5
6
1
2
{1,2}
{5,6}
Summary graph size + Information loss
Merge (1, {5,6})
Merge (1, 2)
Merge (1, 3)
Merge (1, 3)
Graph Summarization is
A Search Problem
How to choose a next action?
Main Idea: MDL Principle
Introduction Algorithms Experiments ConclusionProblem
1
2
3
4
5
6
Merge (5, 6)
1
2
3
4
Merge (1, 2)
3
4
5
6
5
6
1
2
{1,2}
{5,6}
Summary graph size + Information loss
MDL Principle
Merge (1, {5,6})
Merge (1, 2)
Merge (1, 3)
Merge (1, 3)
Graph Summarization is
A Search Problem
How to choose a next action?
arg min 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 �𝑮𝑮 + 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑮𝑮|�𝑮𝑮)
�𝑮𝑮
# 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑓𝑓𝑓𝑓𝑓𝑓 �𝑮𝑮 # 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑓𝑓𝑓𝑓𝑓𝑓 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑮𝑮
𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 �𝑮𝑮
Overview: SSumM
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
Introduction Algorithms Experiments ConclusionProblem
Procedure
• Given:
◦ (1) An input graph 𝑮𝑮, (2) the desired size 𝑲𝑲, (3) the number 𝑻𝑻 of iterations
• Outputs:
◦ Summary graph �𝑮𝑮
Input graph 𝑮𝑮 Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase <<
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
𝑎𝑎
𝑏𝑏𝑐𝑐
𝑑𝑑
𝑒𝑒
𝑓𝑓
𝑔𝑔
ℎ
𝑖𝑖
𝐴𝐴 = {𝑎𝑎}
𝐵𝐵 = {𝑏𝑏}C = {𝑐𝑐}
𝐷𝐷 = {𝑑𝑑}
𝐸𝐸 = {𝑒𝑒}
𝐹𝐹 = {𝑓𝑓}
𝐺𝐺 = {𝑔𝑔}
𝐻𝐻 = {ℎ}
𝐼𝐼 = {𝑖𝑖}
Initialization Phase
Candidate Generation Phase
𝐵𝐵 = {𝑏𝑏}
𝐶𝐶 = {𝑐𝑐}
D = {𝑑𝑑}
𝐸𝐸 = {𝑒𝑒}
𝐴𝐴 = {𝑎𝑎}
𝐹𝐹 = {𝑓𝑓}
𝐺𝐺 = {𝑔𝑔}
𝐻𝐻 = {ℎ}
𝐼𝐼 = {𝑖𝑖}
Input graph 𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase <<
 Merge and sparsification phase
 Further sparsification phase
𝑎𝑎
𝑏𝑏𝑐𝑐
𝑑𝑑
𝑒𝑒
𝑓𝑓
𝑔𝑔
ℎ
𝑖𝑖
Merging and Sparsification Phase
For each candidate set 𝑪𝑪 Among possible candidate pairs
Introduction Algorithms Experiments ConclusionProblem
(A, B) (B, C) (C, D) (D, E)
(A, C) (B, D) (C, E)
(A, D) (B, E)
(A, E)
𝐵𝐵 = {𝑏𝑏}
𝐶𝐶 = {𝑐𝑐}
D = {𝑑𝑑}
𝐴𝐴 = {𝑎𝑎}
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
𝐸𝐸 = {𝑒𝑒}
Merging and Sparsification Phase
For each candidate set 𝑪𝑪 Among possible candidate pairs
Introduction Algorithms Experiments ConclusionProblem
(A, B) (B, C) (C, D) (D, E)
(A, C) (B, D) (C, E)
(A, D) (B, E)
(A, E)
𝐵𝐵 = {𝑏𝑏}
𝐶𝐶 = {𝑐𝑐}
D = {𝑑𝑑}
𝐴𝐴 = {𝑎𝑎}
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
Sample log2 |𝑪𝑪| pairs𝐸𝐸 = {𝑒𝑒}
Merging and Sparsification Phase
Select the pair with
the greatest (relative) reduction
in the cost function
𝒊𝒊𝒊𝒊 reduction(C, D) > 𝜽𝜽:
merge(C, D)
𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆
sample log2 |𝑪𝑪| pairs again
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
(A, B) (A, D) (C, D)
Merging and Sparsification Phase
Select the pair with
the greatest (relative) reduction
in the cost function
𝒊𝒊𝒊𝒊 reduction(C, D) > 𝜽𝜽:
merge(C, D)
𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆
sample log2 |𝑪𝑪| pairs again
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
(A, B) (A, D) (C, D)
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}𝐶𝐶 = {𝑐𝑐}
𝐷𝐷 = {𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
C = {𝑐𝑐, 𝑑𝑑}
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
C = {𝑐𝑐, 𝑑𝑑}
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
Sparsify or not according
to total description cost
{𝑎𝑎}
{𝑏𝑏}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
C = {𝑐𝑐, 𝑑𝑑}
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
Sparsify or not according
to total description cost
{𝑎𝑎}
{𝑏𝑏}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
C = {𝑐𝑐, 𝑑𝑑}
Merging and Sparsification Phase (cont.)
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase <<
 Further sparsification phase
Sparsify or not according
to total description cost
{𝑎𝑎}
{𝑏𝑏}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
C = {𝑐𝑐, 𝑑𝑑}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {ℎ}
{𝑔𝑔, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {ℎ}
{𝑔𝑔, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {ℎ}
{𝑔𝑔, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔, ℎ, 𝑖𝑖}
Summary graph �𝑮𝑮
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑓𝑓}
{𝑔𝑔, ℎ, 𝑖𝑖}
Summary graph �𝑮𝑮
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
{𝑎𝑎, 𝑒𝑒}
Repetition
Summary graph �𝑮𝑮
Introduction Algorithms Experiments ConclusionProblem
Summary graph �𝑮𝑮
Different
candidate sets
and decreasing
threshold 𝜃𝜃
over iteration
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑓𝑓}
{𝑔𝑔, ℎ, 𝑖𝑖}
Summary graph �𝑮𝑮
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖}
{𝑎𝑎}
{𝑏𝑏}
{𝑐𝑐, 𝑑𝑑}
{𝑒𝑒}
{𝑓𝑓}
{𝑔𝑔}
{ℎ}
{𝑖𝑖}
{𝑎𝑎, 𝑒𝑒}
Introduction Algorithms Experiments ConclusionProblem
Further Sparsification Phase
Summary graph �𝑮𝑮
𝐴𝐴 𝐴𝐴
𝐵𝐵 𝐴𝐴
𝐹𝐹 𝐴𝐴
𝐺𝐺 𝐺𝐺
Superedges sorted by ∆𝑅𝑅𝐸𝐸𝑝𝑝
𝐶𝐶 𝐶𝐶
Size of �𝑮𝑮 in bits ≤ 𝑲𝑲
Procedure
 Initialization phase
 𝑡𝑡 = 1
 While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits
 Candidate generation phase
 Merge and sparsification phase
 Further sparsification phase <<
C = {𝑐𝑐, 𝑑𝑑}
𝐴𝐴 = {𝑎𝑎, 𝑒𝑒}
𝐵𝐵 = {𝑏𝑏}
𝐹𝐹 = {𝑓𝑓}
𝐺𝐺 = {𝑔𝑔, ℎ, 𝑖𝑖}
Road Map
• Introduction
• Problem
• Proposed Algorithm: SSumM
• Experimental Results <<
• Conclusions
• 10 datasets from 6 domains (up to 0.8B edges)
• Three competitors for graph summarization
◦ k-Gs [LT10]
◦ S2L [RSB17]
◦ SAA-Gs [BAZK18]
Introduction Algorithms Experiments ConclusionProblem
Social Internet Email Co-purchase Collaboration Hyperlinks
Experiments Settings
Email-Enron Caida Ego-Facebook
Web-UK-05 Web-UK-02 LiveJournal
DBLP Amazon-0302Skitter
k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample)
Introduction Algorithms Experiments ConclusionProblem
o.o.t. >12hours
o.o.m. >64GB
SSumM Gives Concise and Accurate Summary
Email-Enron Caida Ego-Facebook
Web-UK-05 Web-UK-02 LiveJournal
DBLP Amazon-0302Skitter
k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample)
Introduction Algorithms Experiments ConclusionProblem
o.o.t. >12hours
o.o.m. >64GB
SSumM Gives Concise and Accurate Summary
Email-Enron Caida Ego-Facebook
Web-UK-05 Web-UK-02 LiveJournal
Amazon-0601 DBLPSkitter
k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample)
Introduction Algorithms Experiments ConclusionProblem
SSumM is Fast
Email-Enron Caida Ego-Facebook
Web-UK-05 Web-UK-02 LiveJournal
Amazon-0601 DBLPSkitter
k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample)
Introduction Algorithms Experiments ConclusionProblem
SSumM is Fast
Introduction Algorithms Experiments ConclusionProblem
SSumM is Scalable
Introduction Algorithms Experiments ConclusionProblem
SSumM Converges Fast
Road Map
• Introduction
• Problem
• Proposed Algorithm: SSumM
• Experimental Results
• Conclusions <<
Code available at https://github.com/KyuhanLee/SSumM
Concise & Accurate Fast Scalable
Introduction Algorithms Experiments ConclusionProblem
Practical Problem Formulation
Extensive Experiments on 10 real world graphs
Scalable and Effective Algorithm Design
Conclusions
SSumM : Sparse Summarization
of Massive Graphs
Kyuhan Lee* Hyeonsoo Jo* Jihoon Ko Sungsu Lim Kijung Shin

Contenu connexe

Tendances

Image Classification And Support Vector Machine
Image Classification And Support Vector MachineImage Classification And Support Vector Machine
Image Classification And Support Vector MachineShao-Chuan Wang
 
SVM Tutorial
SVM TutorialSVM Tutorial
SVM Tutorialbutest
 
A survey of deep learning approaches to medical applications
A survey of deep learning approaches to medical applicationsA survey of deep learning approaches to medical applications
A survey of deep learning approaches to medical applicationsJoseph Paul Cohen PhD
 
Dimensionality reduction with UMAP
Dimensionality reduction with UMAPDimensionality reduction with UMAP
Dimensionality reduction with UMAPJakub Bartczuk
 
Computer Vision with Deep Learning
Computer Vision with Deep LearningComputer Vision with Deep Learning
Computer Vision with Deep LearningCapgemini
 
Kernels in convolution
Kernels in convolutionKernels in convolution
Kernels in convolutionRevanth Kumar
 
Introduction to object detection
Introduction to object detectionIntroduction to object detection
Introduction to object detectionAmar Jindal
 
Introduction to Deep learning
Introduction to Deep learningIntroduction to Deep learning
Introduction to Deep learningleopauly
 
Neural Networks: Principal Component Analysis (PCA)
Neural Networks: Principal Component Analysis (PCA)Neural Networks: Principal Component Analysis (PCA)
Neural Networks: Principal Component Analysis (PCA)Mostafa G. M. Mostafa
 
K-means Clustering with Scikit-Learn
K-means Clustering with Scikit-LearnK-means Clustering with Scikit-Learn
K-means Clustering with Scikit-LearnSarah Guido
 
Regularization in deep learning
Regularization in deep learningRegularization in deep learning
Regularization in deep learningKien Le
 
Deep Learning - Convolutional Neural Networks
Deep Learning - Convolutional Neural NetworksDeep Learning - Convolutional Neural Networks
Deep Learning - Convolutional Neural NetworksChristian Perone
 
Overview on Optimization algorithms in Deep Learning
Overview on Optimization algorithms in Deep LearningOverview on Optimization algorithms in Deep Learning
Overview on Optimization algorithms in Deep LearningKhang Pham
 
Numerical analysis canonical
Numerical analysis  canonicalNumerical analysis  canonical
Numerical analysis canonicalSHAMJITH KM
 
Controlled dropout: a different dropout for improving training speed on deep ...
Controlled dropout: a different dropout for improving training speed on deep ...Controlled dropout: a different dropout for improving training speed on deep ...
Controlled dropout: a different dropout for improving training speed on deep ...Byung Soo Ko
 

Tendances (20)

GoogLeNet Insights
GoogLeNet InsightsGoogLeNet Insights
GoogLeNet Insights
 
Image Classification And Support Vector Machine
Image Classification And Support Vector MachineImage Classification And Support Vector Machine
Image Classification And Support Vector Machine
 
SVM Tutorial
SVM TutorialSVM Tutorial
SVM Tutorial
 
A survey of deep learning approaches to medical applications
A survey of deep learning approaches to medical applicationsA survey of deep learning approaches to medical applications
A survey of deep learning approaches to medical applications
 
Dimensionality reduction with UMAP
Dimensionality reduction with UMAPDimensionality reduction with UMAP
Dimensionality reduction with UMAP
 
Computer Vision with Deep Learning
Computer Vision with Deep LearningComputer Vision with Deep Learning
Computer Vision with Deep Learning
 
CNN
CNNCNN
CNN
 
Kernels in convolution
Kernels in convolutionKernels in convolution
Kernels in convolution
 
Perceptron in ANN
Perceptron in ANNPerceptron in ANN
Perceptron in ANN
 
Introduction to object detection
Introduction to object detectionIntroduction to object detection
Introduction to object detection
 
Introduction to Deep learning
Introduction to Deep learningIntroduction to Deep learning
Introduction to Deep learning
 
svm classification
svm classificationsvm classification
svm classification
 
CNN Tutorial
CNN TutorialCNN Tutorial
CNN Tutorial
 
Neural Networks: Principal Component Analysis (PCA)
Neural Networks: Principal Component Analysis (PCA)Neural Networks: Principal Component Analysis (PCA)
Neural Networks: Principal Component Analysis (PCA)
 
K-means Clustering with Scikit-Learn
K-means Clustering with Scikit-LearnK-means Clustering with Scikit-Learn
K-means Clustering with Scikit-Learn
 
Regularization in deep learning
Regularization in deep learningRegularization in deep learning
Regularization in deep learning
 
Deep Learning - Convolutional Neural Networks
Deep Learning - Convolutional Neural NetworksDeep Learning - Convolutional Neural Networks
Deep Learning - Convolutional Neural Networks
 
Overview on Optimization algorithms in Deep Learning
Overview on Optimization algorithms in Deep LearningOverview on Optimization algorithms in Deep Learning
Overview on Optimization algorithms in Deep Learning
 
Numerical analysis canonical
Numerical analysis  canonicalNumerical analysis  canonical
Numerical analysis canonical
 
Controlled dropout: a different dropout for improving training speed on deep ...
Controlled dropout: a different dropout for improving training speed on deep ...Controlled dropout: a different dropout for improving training speed on deep ...
Controlled dropout: a different dropout for improving training speed on deep ...
 

Similaire à "SSumM: Sparse Summarization of Massive Graphs", KDD 2020

Application of parallel hierarchical matrices for parameter inference and pre...
Application of parallel hierarchical matrices for parameter inference and pre...Application of parallel hierarchical matrices for parameter inference and pre...
Application of parallel hierarchical matrices for parameter inference and pre...Alexander Litvinenko
 
Identification of unknown parameters and prediction of missing values. Compar...
Identification of unknown parameters and prediction of missing values. Compar...Identification of unknown parameters and prediction of missing values. Compar...
Identification of unknown parameters and prediction of missing values. Compar...Alexander Litvinenko
 
Identification of unknown parameters and prediction with hierarchical matrice...
Identification of unknown parameters and prediction with hierarchical matrice...Identification of unknown parameters and prediction with hierarchical matrice...
Identification of unknown parameters and prediction with hierarchical matrice...Alexander Litvinenko
 
Brief instruction on backprop
Brief instruction on backpropBrief instruction on backprop
Brief instruction on backpropYasutoTamura1
 
Pratt truss optimization using
Pratt truss optimization usingPratt truss optimization using
Pratt truss optimization usingHarish Kant Soni
 
Enumerating cycles in bipartite graph using matrix approach
Enumerating cycles in bipartite graph using matrix approachEnumerating cycles in bipartite graph using matrix approach
Enumerating cycles in bipartite graph using matrix approachUsatyuk Vasiliy
 
Nelson maple pdf
Nelson maple pdfNelson maple pdf
Nelson maple pdfNelsonP23
 
Sample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docx
Sample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docxSample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docx
Sample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docxanhlodge
 
Measurement of reliability parameters for a power
Measurement of reliability parameters for a powerMeasurement of reliability parameters for a power
Measurement of reliability parameters for a powereSAT Publishing House
 
Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques Alexander Litvinenko
 
PREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial Approach
PREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial ApproachPREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial Approach
PREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial Approachahmet furkan emrehan
 
Computer Vision: Correlation, Convolution, and Gradient
Computer Vision: Correlation, Convolution, and GradientComputer Vision: Correlation, Convolution, and Gradient
Computer Vision: Correlation, Convolution, and GradientAhmed Gad
 
Statistical quality control, sampling
Statistical quality control, samplingStatistical quality control, sampling
Statistical quality control, samplingSana Fatima
 
Shmoo Quantify
Shmoo QuantifyShmoo Quantify
Shmoo QuantifyDan Shu
 

Similaire à "SSumM: Sparse Summarization of Massive Graphs", KDD 2020 (20)

Application of parallel hierarchical matrices for parameter inference and pre...
Application of parallel hierarchical matrices for parameter inference and pre...Application of parallel hierarchical matrices for parameter inference and pre...
Application of parallel hierarchical matrices for parameter inference and pre...
 
Identification of unknown parameters and prediction of missing values. Compar...
Identification of unknown parameters and prediction of missing values. Compar...Identification of unknown parameters and prediction of missing values. Compar...
Identification of unknown parameters and prediction of missing values. Compar...
 
Control charts
Control chartsControl charts
Control charts
 
Identification of unknown parameters and prediction with hierarchical matrice...
Identification of unknown parameters and prediction with hierarchical matrice...Identification of unknown parameters and prediction with hierarchical matrice...
Identification of unknown parameters and prediction with hierarchical matrice...
 
Brief instruction on backprop
Brief instruction on backpropBrief instruction on backprop
Brief instruction on backprop
 
Data Analysis Assignment Help
Data Analysis Assignment HelpData Analysis Assignment Help
Data Analysis Assignment Help
 
icdm06-tong.ppt
icdm06-tong.ppticdm06-tong.ppt
icdm06-tong.ppt
 
Pratt truss optimization using
Pratt truss optimization usingPratt truss optimization using
Pratt truss optimization using
 
Enumerating cycles in bipartite graph using matrix approach
Enumerating cycles in bipartite graph using matrix approachEnumerating cycles in bipartite graph using matrix approach
Enumerating cycles in bipartite graph using matrix approach
 
Aocr Hmm Presentation
Aocr Hmm PresentationAocr Hmm Presentation
Aocr Hmm Presentation
 
Nelson maple pdf
Nelson maple pdfNelson maple pdf
Nelson maple pdf
 
Sample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docx
Sample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docxSample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docx
Sample Data for Problem 1aA =[1 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0.docx
 
Measurement of reliability parameters for a power
Measurement of reliability parameters for a powerMeasurement of reliability parameters for a power
Measurement of reliability parameters for a power
 
Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques
 
PREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial Approach
PREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial ApproachPREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial Approach
PREDICTION MODELS BASED ON MAX-STEMS Episode Two: Combinatorial Approach
 
Computer Vision: Correlation, Convolution, and Gradient
Computer Vision: Correlation, Convolution, and GradientComputer Vision: Correlation, Convolution, and Gradient
Computer Vision: Correlation, Convolution, and Gradient
 
Cluto presentation
Cluto presentationCluto presentation
Cluto presentation
 
Vine shortest example
Vine shortest exampleVine shortest example
Vine shortest example
 
Statistical quality control, sampling
Statistical quality control, samplingStatistical quality control, sampling
Statistical quality control, sampling
 
Shmoo Quantify
Shmoo QuantifyShmoo Quantify
Shmoo Quantify
 

Dernier

YourView Panel Book.pptx YourView Panel Book.
YourView Panel Book.pptx YourView Panel Book.YourView Panel Book.pptx YourView Panel Book.
YourView Panel Book.pptx YourView Panel Book.JasonViviers2
 
Master's Thesis - Data Science - Presentation
Master's Thesis - Data Science - PresentationMaster's Thesis - Data Science - Presentation
Master's Thesis - Data Science - PresentationGiorgio Carbone
 
Strategic CX: A Deep Dive into Voice of the Customer Insights for Clarity
Strategic CX: A Deep Dive into Voice of the Customer Insights for ClarityStrategic CX: A Deep Dive into Voice of the Customer Insights for Clarity
Strategic CX: A Deep Dive into Voice of the Customer Insights for ClarityAggregage
 
SFBA Splunk Usergroup meeting March 13, 2024
SFBA Splunk Usergroup meeting March 13, 2024SFBA Splunk Usergroup meeting March 13, 2024
SFBA Splunk Usergroup meeting March 13, 2024Becky Burwell
 
Cash Is Still King: ATM market research '2023
Cash Is Still King: ATM market research '2023Cash Is Still King: ATM market research '2023
Cash Is Still King: ATM market research '2023Vladislav Solodkiy
 
TINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptx
TINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptxTINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptx
TINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptxDwiAyuSitiHartinah
 
Elements of language learning - an analysis of how different elements of lang...
Elements of language learning - an analysis of how different elements of lang...Elements of language learning - an analysis of how different elements of lang...
Elements of language learning - an analysis of how different elements of lang...PrithaVashisht1
 
ChistaDATA Real-Time DATA Analytics Infrastructure
ChistaDATA Real-Time DATA Analytics InfrastructureChistaDATA Real-Time DATA Analytics Infrastructure
ChistaDATA Real-Time DATA Analytics Infrastructuresonikadigital1
 
Virtuosoft SmartSync Product Introduction
Virtuosoft SmartSync Product IntroductionVirtuosoft SmartSync Product Introduction
Virtuosoft SmartSync Product Introductionsanjaymuralee1
 
Mapping the pubmed data under different suptopics using NLP.pptx
Mapping the pubmed data under different suptopics using NLP.pptxMapping the pubmed data under different suptopics using NLP.pptx
Mapping the pubmed data under different suptopics using NLP.pptxVenkatasubramani13
 
How is Real-Time Analytics Different from Traditional OLAP?
How is Real-Time Analytics Different from Traditional OLAP?How is Real-Time Analytics Different from Traditional OLAP?
How is Real-Time Analytics Different from Traditional OLAP?sonikadigital1
 
CI, CD -Tools to integrate without manual intervention
CI, CD -Tools to integrate without manual interventionCI, CD -Tools to integrate without manual intervention
CI, CD -Tools to integrate without manual interventionajayrajaganeshkayala
 
AI for Sustainable Development Goals (SDGs)
AI for Sustainable Development Goals (SDGs)AI for Sustainable Development Goals (SDGs)
AI for Sustainable Development Goals (SDGs)Data & Analytics Magazin
 
Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024
Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024
Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024Guido X Jansen
 
MEASURES OF DISPERSION I BSc Botany .ppt
MEASURES OF DISPERSION I BSc Botany .pptMEASURES OF DISPERSION I BSc Botany .ppt
MEASURES OF DISPERSION I BSc Botany .pptaigil2
 
The Universal GTM - how we design GTM and dataLayer
The Universal GTM - how we design GTM and dataLayerThe Universal GTM - how we design GTM and dataLayer
The Universal GTM - how we design GTM and dataLayerPavel Šabatka
 
5 Ds to Define Data Archiving Best Practices
5 Ds to Define Data Archiving Best Practices5 Ds to Define Data Archiving Best Practices
5 Ds to Define Data Archiving Best PracticesDataArchiva
 

Dernier (17)

YourView Panel Book.pptx YourView Panel Book.
YourView Panel Book.pptx YourView Panel Book.YourView Panel Book.pptx YourView Panel Book.
YourView Panel Book.pptx YourView Panel Book.
 
Master's Thesis - Data Science - Presentation
Master's Thesis - Data Science - PresentationMaster's Thesis - Data Science - Presentation
Master's Thesis - Data Science - Presentation
 
Strategic CX: A Deep Dive into Voice of the Customer Insights for Clarity
Strategic CX: A Deep Dive into Voice of the Customer Insights for ClarityStrategic CX: A Deep Dive into Voice of the Customer Insights for Clarity
Strategic CX: A Deep Dive into Voice of the Customer Insights for Clarity
 
SFBA Splunk Usergroup meeting March 13, 2024
SFBA Splunk Usergroup meeting March 13, 2024SFBA Splunk Usergroup meeting March 13, 2024
SFBA Splunk Usergroup meeting March 13, 2024
 
Cash Is Still King: ATM market research '2023
Cash Is Still King: ATM market research '2023Cash Is Still King: ATM market research '2023
Cash Is Still King: ATM market research '2023
 
TINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptx
TINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptxTINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptx
TINJUAN PEMROSESAN TRANSAKSI DAN ERP.pptx
 
Elements of language learning - an analysis of how different elements of lang...
Elements of language learning - an analysis of how different elements of lang...Elements of language learning - an analysis of how different elements of lang...
Elements of language learning - an analysis of how different elements of lang...
 
ChistaDATA Real-Time DATA Analytics Infrastructure
ChistaDATA Real-Time DATA Analytics InfrastructureChistaDATA Real-Time DATA Analytics Infrastructure
ChistaDATA Real-Time DATA Analytics Infrastructure
 
Virtuosoft SmartSync Product Introduction
Virtuosoft SmartSync Product IntroductionVirtuosoft SmartSync Product Introduction
Virtuosoft SmartSync Product Introduction
 
Mapping the pubmed data under different suptopics using NLP.pptx
Mapping the pubmed data under different suptopics using NLP.pptxMapping the pubmed data under different suptopics using NLP.pptx
Mapping the pubmed data under different suptopics using NLP.pptx
 
How is Real-Time Analytics Different from Traditional OLAP?
How is Real-Time Analytics Different from Traditional OLAP?How is Real-Time Analytics Different from Traditional OLAP?
How is Real-Time Analytics Different from Traditional OLAP?
 
CI, CD -Tools to integrate without manual intervention
CI, CD -Tools to integrate without manual interventionCI, CD -Tools to integrate without manual intervention
CI, CD -Tools to integrate without manual intervention
 
AI for Sustainable Development Goals (SDGs)
AI for Sustainable Development Goals (SDGs)AI for Sustainable Development Goals (SDGs)
AI for Sustainable Development Goals (SDGs)
 
Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024
Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024
Persuasive E-commerce, Our Biased Brain @ Bikkeldag 2024
 
MEASURES OF DISPERSION I BSc Botany .ppt
MEASURES OF DISPERSION I BSc Botany .pptMEASURES OF DISPERSION I BSc Botany .ppt
MEASURES OF DISPERSION I BSc Botany .ppt
 
The Universal GTM - how we design GTM and dataLayer
The Universal GTM - how we design GTM and dataLayerThe Universal GTM - how we design GTM and dataLayer
The Universal GTM - how we design GTM and dataLayer
 
5 Ds to Define Data Archiving Best Practices
5 Ds to Define Data Archiving Best Practices5 Ds to Define Data Archiving Best Practices
5 Ds to Define Data Archiving Best Practices
 

"SSumM: Sparse Summarization of Massive Graphs", KDD 2020

  • 1. SSumM : Sparse Summarization of Massive Graphs Kyuhan Lee* Hyeonsoo Jo* Jihoon Ko Sungsu Lim Kijung Shin
  • 2. Graphs are Everywhere Citation networksSubway networks Internet topologies Introduction Algorithms Experiments ConclusionProblem Designed by Freepik from Flaticon
  • 3. Massive Graphs Appeared Social networks 2.49 Billion active users Purchase histories 0.5 Billion products World Wide Web 5.49 Billion web pages Introduction Algorithms Experiments ConclusionProblem
  • 4. Difficulties in Analyzing Massive graphs Computational cost (number of nodes & edges) Introduction Algorithms Experiments ConclusionProblem
  • 5. Difficulties in Analyzing Massive graphs > Cannot fit Introduction Algorithms Experiments ConclusionProblem Input Graph
  • 6. Solution: Graph Summarization > Introduction Algorithms Experiments ConclusionProblem Can fit Summary Graph
  • 7. Advantages of Graph Summarization Introduction Algorithms Experiments ConclusionProblem • Many graph compression techniques are available • TheWebGraph Framework [BV04] • BFS encoding [AD09] • SlashBurn [KF11] • VoG [KKVF14] • Graph summarization stands out because • Elastic: reduce size of outputs as much as we want • Analyzable: existing graph analysis and tools can be applied • Combinable for Additional Compression: can be further compressed
  • 8. Example of Graph Summarization 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Adjacency Matrix 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Introduction Algorithms Experiments ConclusionProblem Input Graph
  • 9. Example of Graph Summarization 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Adjacency Matrix Input Graph 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 𝑽𝑽𝟏𝟏={1,2} 𝑽𝑽𝟕𝟕={7,8,9} 𝑽𝑽𝟑𝟑={3,4,5,6} Summary Graph Introduction Algorithms Experiments ConclusionProblem Subnode Subedge Supernode
  • 10. Example of Graph Summarization 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Adjacency Matrix Input Graph 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 𝑽𝑽𝟏𝟏={1,2} 𝑽𝑽𝟕𝟕={7,8,9} 𝑽𝑽𝟑𝟑={3,4,5,6} Summary Graph 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3 Introduction Algorithms Experiments ConclusionProblem Subnode Subedge Superedge Supernode
  • 11. Example of Graph Summarization 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Adjacency Matrix Input Graph 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 𝑽𝑽𝟏𝟏={1,2} 𝑽𝑽𝟕𝟕={7,8,9} 𝑽𝑽𝟑𝟑={3,4,5,6} Summary Graph 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3 𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟕𝟕} =2 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1 𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3 𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5 Introduction Algorithms Experiments ConclusionProblem
  • 12. Example of Graph Summarization 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Adjacency Matrix Input Graph 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 𝑽𝑽𝟏𝟏={1,2} 𝑽𝑽𝟕𝟕={7,8,9} 𝑽𝑽𝟑𝟑={3,4,5,6} Summary Graph 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1 𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3 𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5 Introduction Algorithms Experiments ConclusionProblem
  • 13. Example of Graph Summarization Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 7 8 9 1 0 1 3/8 3/8 3/8 3/8 1/3 1/3 1/3 2 1 0 3/8 3/8 3/8 3/8 1/3 1/3 1/3 3 3/8 3/8 0 5/6 5/6 5/6 0 0 0 4 3/8 3/8 5/6 0 5/6 5/6 0 0 0 5 3/8 3/8 5/6 5/6 0 5/6 0 0 0 6 3/8 3/8 5/6 5/6 5/6 0 0 0 0 7 1/3 1/3 0 0 0 0 0 1 1 8 1/3 1/3 0 0 0 0 1 0 1 9 1/3 1/3 0 0 0 0 1 1 0 Reconstructed Adjacency Matrix 𝑽𝑽𝟏𝟏={1,2} 𝑽𝑽𝟕𝟕={7,8,9} 𝑽𝑽𝟑𝟑={3,4,5,6} Summary Graph 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1 𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3 𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5
  • 14. Example of Graph Summarization 𝑽𝑽𝟏𝟏={1,2} 𝑽𝑽𝟕𝟕={7,8,9} 𝑽𝑽𝟑𝟑={3,4,5,6} Summary Graph 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟕𝟕} =2 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} =3𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟏𝟏} =1 𝝎𝝎 {𝑽𝑽𝟕𝟕, 𝑽𝑽𝟕𝟕} =3 𝝎𝝎 {𝑽𝑽𝟑𝟑, 𝑽𝑽𝟑𝟑} =5 Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 7 8 9 1 0 1 3/8 3/8 3/8 3/8 1/3 1/3 1/3 2 1 0 3/8 3/8 3/8 3/8 1/3 1/3 1/3 3 3/8 3/8 0 5/6 5/6 5/6 0 0 0 4 3/8 3/8 5/6 0 5/6 5/6 0 0 0 5 3/8 3/8 5/6 5/6 0 5/6 0 0 0 6 3/8 3/8 5/6 5/6 5/6 0 0 0 0 7 1/3 1/3 0 0 0 0 0 1 1 8 1/3 1/3 0 0 0 0 1 0 1 9 1/3 1/3 0 0 0 0 1 1 0 Reconstructed Adjacency Matrix 𝝎𝝎 {𝑽𝑽𝟏𝟏, 𝑽𝑽𝟑𝟑} = 3 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀𝑀𝑀 𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛𝑛 𝑜𝑜𝑜𝑜 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 8
  • 15. Road Map • Introduction • Problem << • Proposed Algorithm: SSumM • Experimental Results • Conclusions
  • 16. Problem Definition: Graph Summarization 𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮: a graph 𝑮𝑮 and the target number of node 𝑲𝑲 𝑭𝑭𝑭𝑭𝑭𝑭𝑭𝑭: a summary graph �𝑮𝑮 𝑻𝑻𝑻𝑻 𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴: the difference between graph 𝑮𝑮and the restored graph �𝑮𝑮 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒕𝒕𝒕𝒕: the number of supernodes in �𝑮𝑮 ≤ 𝑲𝑲 Introduction Algorithms Experiments ConclusionProblem
  • 17. Problem Definition: Graph Summarization 𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮: a graph 𝑮𝑮 and the target number of node 𝑲𝑲 𝑭𝑭𝑭𝑭𝑭𝑭𝑭𝑭: a summary graph �𝑮𝑮 𝑻𝑻𝑻𝑻 𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴: the difference between graph 𝑮𝑮and the restored graph �𝑮𝑮 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒕𝒕𝒕𝒕: the number of supernodes in �𝑮𝑮 ≤ 𝑲𝑲 Shouldn’t we consider sizes? Introduction Algorithms Experiments ConclusionProblem
  • 18. Problem Definition: Graph Summarization 𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮𝑮: a graph 𝑮𝑮 and the desired size 𝑲𝑲 (in bits) 𝑭𝑭𝑭𝑭𝑭𝑭𝑭𝑭: a summary graph �𝑮𝑮 𝑻𝑻𝑻𝑻 𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴𝑴: the difference with graph graph 𝑮𝑮and the restored graph �𝑮𝑮 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒕𝒕𝒕𝒕: size of �𝑮𝑮 in bits ≤ 𝑲𝑲 Introduction Algorithms Experiments ConclusionProblem
  • 19. Details: Size in Bits of a Graph 𝐸𝐸 : set of edges 𝑉𝑉 : set of nodes Encoded using log2|𝑉𝑉| bits Introduction Algorithms Experiments ConclusionProblem 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 2 𝐸𝐸 log2 𝑉𝑉 Input graph 𝑮𝑮
  • 20. Details: Size in Bits of a Summary Graph 4 5 1 4 1 5 Introduction Algorithms Experiments ConclusionProblem 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆 𝑆𝑆 : set of supernodes 𝑃𝑃 : set of superedges 𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight Summary graph �𝑮𝑮
  • 21. Details: Size in Bits of a Summary Graph 4 5 1 4 1 5 Introduction Algorithms Experiments ConclusionProblem 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆 𝑆𝑆 : set of supernodes 𝑃𝑃 : set of superedges 𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight Summary graph �𝑮𝑮
  • 22. Details: Size in Bits of a Summary Graph 4 5 1 4 1 5 Introduction Algorithms Experiments ConclusionProblem 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆 𝑆𝑆 : set of supernodes 𝑃𝑃 : set of superedges 𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight Summary graph �𝑮𝑮
  • 23. Details: Size in Bits of a Summary Graph 4 5 1 4 1 5 Introduction Algorithms Experiments ConclusionProblem 𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝒐𝒐𝒐𝒐 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈𝒈: 𝑃𝑃 2 log2 𝑆𝑆 + log2 𝜔𝜔𝑚𝑚𝑚𝑚𝑚𝑚 + 𝑉𝑉 log2 𝑆𝑆 𝑆𝑆 : set of supernodes 𝑃𝑃 : set of superedges 𝑤𝑤𝑚𝑚𝑚𝑚𝑚𝑚 : maximum superedge weight Summary graph �𝑮𝑮
  • 24. Details: Error Measurement Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 7 8 9 1 0 1 3/8 3/8 3/8 3/8 1/3 1/3 1/3 2 1 0 3/8 3/8 3/8 3/8 1/3 1/3 1/3 3 3/8 3/8 0 5/6 5/6 5/6 0 0 0 4 3/8 3/8 5/6 0 5/6 5/6 0 0 0 5 3/8 3/8 5/6 5/6 0 5/6 0 0 0 6 3/8 3/8 5/6 5/6 5/6 0 0 0 0 7 1/3 1/3 0 0 0 0 0 1 1 8 1/3 1/3 0 0 0 0 1 0 1 9 1/3 1/3 0 0 0 0 1 1 0 1 2 3 4 5 6 7 8 9 1 0 1 0 0 0 1 0 0 1 2 1 0 1 0 0 1 1 0 0 3 0 1 0 1 1 1 0 0 1 4 0 0 1 0 1 0 0 0 0 5 0 0 1 1 0 1 1 0 0 6 1 1 1 0 1 0 0 0 0 7 0 1 0 0 1 0 0 1 1 8 0 0 0 0 0 0 1 0 1 9 1 0 1 0 0 0 1 1 0 Reconstructed Adjacency Matrix �𝑨𝑨Reconstructed Adjacency Matrix 𝑨𝑨 𝑅𝑅𝐸𝐸𝑝𝑝(𝑨𝑨, �𝑨𝑨) = � 𝑖𝑖=1 𝑉𝑉 � 𝑗𝑗=1 𝑉𝑉 𝐴𝐴 𝑖𝑖, 𝑗𝑗 − ̂𝐴𝐴 𝑖𝑖, 𝑗𝑗 𝑝𝑝 𝟏𝟏 𝒑𝒑
  • 25. Road Map • Introduction • Problem • Proposed Algorithm: SSumM << • Experimental Results • Conclusions
  • 26. Main ideas of SSumM Introduction Algorithms Experiments ConclusionProblem Combines node grouping and edge sparsification Prunes search space Balances error and size of the summary graph using MDL principle • Practical graph summarization problem ◦ Given: a graph 𝑮𝑮 ◦ Find: a summary graph �𝑮𝑮 ◦ To minimize: the difference between 𝑮𝑮and the restored graph �𝑮𝑮 ◦ Subject to: 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑜𝑜𝑜𝑜 �𝑮𝑮 in bits ≤ 𝑲𝑲
  • 27. Main Idea: Combining Two Strategies Introduction Algorithms Experiments ConclusionProblem Node Grouping Sparsification
  • 28. Main Idea: Combining Two Strategies Introduction Algorithms Experiments ConclusionProblem Node Grouping Sparsification
  • 29. Main Idea: Combining Two Strategies Introduction Algorithms Experiments ConclusionProblem Node Grouping Sparsification
  • 30. Main Idea: Combining Two Strategies Introduction Algorithms Experiments ConclusionProblem Node Grouping Sparsification
  • 31. Main Idea: MDL Principle Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 Merge (5, 6) 1 2 3 4 Merge (1, 2) 3 4 5 6 5 6 1 2 {1,2} {5,6}Merge (1, {5,6}) Merge (1, 2) Merge (1, 3) Merge (1, 3) How to choose a next action?
  • 32. Main Idea: MDL Principle Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 Merge (5, 6) 1 2 3 4 Merge (1, 2) 3 4 5 6 5 6 1 2 {1,2} {5,6}Merge (1, {5,6}) Merge (1, 2) Merge (1, 3) Merge (1, 3) Graph Summarization is A Search Problem How to choose a next action?
  • 33. Main Idea: MDL Principle Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 Merge (5, 6) 1 2 3 4 Merge (1, 2) 3 4 5 6 5 6 1 2 {1,2} {5,6} Summary graph size + Information loss Merge (1, {5,6}) Merge (1, 2) Merge (1, 3) Merge (1, 3) Graph Summarization is A Search Problem How to choose a next action?
  • 34. Main Idea: MDL Principle Introduction Algorithms Experiments ConclusionProblem 1 2 3 4 5 6 Merge (5, 6) 1 2 3 4 Merge (1, 2) 3 4 5 6 5 6 1 2 {1,2} {5,6} Summary graph size + Information loss MDL Principle Merge (1, {5,6}) Merge (1, 2) Merge (1, 3) Merge (1, 3) Graph Summarization is A Search Problem How to choose a next action? arg min 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 �𝑮𝑮 + 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶(𝑮𝑮|�𝑮𝑮) �𝑮𝑮 # 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑓𝑓𝑓𝑓𝑓𝑓 �𝑮𝑮 # 𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑓𝑓𝑓𝑓𝑓𝑓 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝑮𝑮 𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 �𝑮𝑮
  • 35. Overview: SSumM  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase Introduction Algorithms Experiments ConclusionProblem Procedure • Given: ◦ (1) An input graph 𝑮𝑮, (2) the desired size 𝑲𝑲, (3) the number 𝑻𝑻 of iterations • Outputs: ◦ Summary graph �𝑮𝑮
  • 36. Input graph 𝑮𝑮 Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase <<  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase 𝑎𝑎 𝑏𝑏𝑐𝑐 𝑑𝑑 𝑒𝑒 𝑓𝑓 𝑔𝑔 ℎ 𝑖𝑖 𝐴𝐴 = {𝑎𝑎} 𝐵𝐵 = {𝑏𝑏}C = {𝑐𝑐} 𝐷𝐷 = {𝑑𝑑} 𝐸𝐸 = {𝑒𝑒} 𝐹𝐹 = {𝑓𝑓} 𝐺𝐺 = {𝑔𝑔} 𝐻𝐻 = {ℎ} 𝐼𝐼 = {𝑖𝑖} Initialization Phase
  • 37. Candidate Generation Phase 𝐵𝐵 = {𝑏𝑏} 𝐶𝐶 = {𝑐𝑐} D = {𝑑𝑑} 𝐸𝐸 = {𝑒𝑒} 𝐴𝐴 = {𝑎𝑎} 𝐹𝐹 = {𝑓𝑓} 𝐺𝐺 = {𝑔𝑔} 𝐻𝐻 = {ℎ} 𝐼𝐼 = {𝑖𝑖} Input graph 𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase <<  Merge and sparsification phase  Further sparsification phase 𝑎𝑎 𝑏𝑏𝑐𝑐 𝑑𝑑 𝑒𝑒 𝑓𝑓 𝑔𝑔 ℎ 𝑖𝑖
  • 38. Merging and Sparsification Phase For each candidate set 𝑪𝑪 Among possible candidate pairs Introduction Algorithms Experiments ConclusionProblem (A, B) (B, C) (C, D) (D, E) (A, C) (B, D) (C, E) (A, D) (B, E) (A, E) 𝐵𝐵 = {𝑏𝑏} 𝐶𝐶 = {𝑐𝑐} D = {𝑑𝑑} 𝐴𝐴 = {𝑎𝑎} Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase 𝐸𝐸 = {𝑒𝑒}
  • 39. Merging and Sparsification Phase For each candidate set 𝑪𝑪 Among possible candidate pairs Introduction Algorithms Experiments ConclusionProblem (A, B) (B, C) (C, D) (D, E) (A, C) (B, D) (C, E) (A, D) (B, E) (A, E) 𝐵𝐵 = {𝑏𝑏} 𝐶𝐶 = {𝑐𝑐} D = {𝑑𝑑} 𝐴𝐴 = {𝑎𝑎} Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase Sample log2 |𝑪𝑪| pairs𝐸𝐸 = {𝑒𝑒}
  • 40. Merging and Sparsification Phase Select the pair with the greatest (relative) reduction in the cost function 𝒊𝒊𝒊𝒊 reduction(C, D) > 𝜽𝜽: merge(C, D) 𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆 sample log2 |𝑪𝑪| pairs again Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase (A, B) (A, D) (C, D)
  • 41. Merging and Sparsification Phase Select the pair with the greatest (relative) reduction in the cost function 𝒊𝒊𝒊𝒊 reduction(C, D) > 𝜽𝜽: merge(C, D) 𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆 sample log2 |𝑪𝑪| pairs again Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase (A, B) (A, D) (C, D)
  • 42. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase {𝑎𝑎} {𝑏𝑏}𝐶𝐶 = {𝑐𝑐} 𝐷𝐷 = {𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 43. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 44. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} C = {𝑐𝑐, 𝑑𝑑}
  • 45. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} C = {𝑐𝑐, 𝑑𝑑}
  • 46. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase Sparsify or not according to total description cost {𝑎𝑎} {𝑏𝑏} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} C = {𝑐𝑐, 𝑑𝑑}
  • 47. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase Sparsify or not according to total description cost {𝑎𝑎} {𝑏𝑏} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} C = {𝑐𝑐, 𝑑𝑑}
  • 48. Merging and Sparsification Phase (cont.) Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase <<  Further sparsification phase Sparsify or not according to total description cost {𝑎𝑎} {𝑏𝑏} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} C = {𝑐𝑐, 𝑑𝑑}
  • 49. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 50. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 51. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 52. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {ℎ} {𝑔𝑔, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 53. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {ℎ} {𝑔𝑔, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 54. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {ℎ} {𝑔𝑔, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 55. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 56. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 57. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} Summary graph �𝑮𝑮 {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖}
  • 58. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} Summary graph �𝑮𝑮 {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} {𝑎𝑎, 𝑒𝑒}
  • 59. Repetition Summary graph �𝑮𝑮 Introduction Algorithms Experiments ConclusionProblem Summary graph �𝑮𝑮 Different candidate sets and decreasing threshold 𝜃𝜃 over iteration Procedure  Initialization phase  𝑡𝑡 = 1  While 𝒕𝒕++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} Summary graph �𝑮𝑮 {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔, ℎ, 𝑖𝑖} {𝑎𝑎} {𝑏𝑏} {𝑐𝑐, 𝑑𝑑} {𝑒𝑒} {𝑓𝑓} {𝑔𝑔} {ℎ} {𝑖𝑖} {𝑎𝑎, 𝑒𝑒}
  • 60. Introduction Algorithms Experiments ConclusionProblem Further Sparsification Phase Summary graph �𝑮𝑮 𝐴𝐴 𝐴𝐴 𝐵𝐵 𝐴𝐴 𝐹𝐹 𝐴𝐴 𝐺𝐺 𝐺𝐺 Superedges sorted by ∆𝑅𝑅𝐸𝐸𝑝𝑝 𝐶𝐶 𝐶𝐶 Size of �𝑮𝑮 in bits ≤ 𝑲𝑲 Procedure  Initialization phase  𝑡𝑡 = 1  While 𝑡𝑡++ ≤ 𝑻𝑻 and 𝑲𝑲 < size of �𝑮𝑮 in bits  Candidate generation phase  Merge and sparsification phase  Further sparsification phase << C = {𝑐𝑐, 𝑑𝑑} 𝐴𝐴 = {𝑎𝑎, 𝑒𝑒} 𝐵𝐵 = {𝑏𝑏} 𝐹𝐹 = {𝑓𝑓} 𝐺𝐺 = {𝑔𝑔, ℎ, 𝑖𝑖}
  • 61. Road Map • Introduction • Problem • Proposed Algorithm: SSumM • Experimental Results << • Conclusions
  • 62. • 10 datasets from 6 domains (up to 0.8B edges) • Three competitors for graph summarization ◦ k-Gs [LT10] ◦ S2L [RSB17] ◦ SAA-Gs [BAZK18] Introduction Algorithms Experiments ConclusionProblem Social Internet Email Co-purchase Collaboration Hyperlinks Experiments Settings
  • 63. Email-Enron Caida Ego-Facebook Web-UK-05 Web-UK-02 LiveJournal DBLP Amazon-0302Skitter k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample) Introduction Algorithms Experiments ConclusionProblem o.o.t. >12hours o.o.m. >64GB SSumM Gives Concise and Accurate Summary
  • 64. Email-Enron Caida Ego-Facebook Web-UK-05 Web-UK-02 LiveJournal DBLP Amazon-0302Skitter k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample) Introduction Algorithms Experiments ConclusionProblem o.o.t. >12hours o.o.m. >64GB SSumM Gives Concise and Accurate Summary
  • 65. Email-Enron Caida Ego-Facebook Web-UK-05 Web-UK-02 LiveJournal Amazon-0601 DBLPSkitter k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample) Introduction Algorithms Experiments ConclusionProblem SSumM is Fast
  • 66. Email-Enron Caida Ego-Facebook Web-UK-05 Web-UK-02 LiveJournal Amazon-0601 DBLPSkitter k-GsSSumM S2LSAA-Gs SAA-Gs (linear sample) Introduction Algorithms Experiments ConclusionProblem SSumM is Fast
  • 67. Introduction Algorithms Experiments ConclusionProblem SSumM is Scalable
  • 68. Introduction Algorithms Experiments ConclusionProblem SSumM Converges Fast
  • 69. Road Map • Introduction • Problem • Proposed Algorithm: SSumM • Experimental Results • Conclusions <<
  • 70. Code available at https://github.com/KyuhanLee/SSumM Concise & Accurate Fast Scalable Introduction Algorithms Experiments ConclusionProblem Practical Problem Formulation Extensive Experiments on 10 real world graphs Scalable and Effective Algorithm Design Conclusions
  • 71. SSumM : Sparse Summarization of Massive Graphs Kyuhan Lee* Hyeonsoo Jo* Jihoon Ko Sungsu Lim Kijung Shin