1οΈβ£ 12κ° λ³΅μ΅
πΉ Main Topic : Subgraph Mining - Identifying and Counting Motfits in Networks
• Subgraph μ motifs λ κ·Έλν ꡬ쑰μ λν μΈμ¬μ΄νΈλ₯Ό μ»κ² ν΄μ£Όλ μμμ΄λ€.
• νΉμ κ·Έλνλ€ μ¬μ΄μμ Subgraph Isomorphism κ΄κ³μ μλμ§ μμΈ‘νλ Task μ λν΄ μ κ²½λ§μ νμ©νλ€.
• Subgraph κ΄κ³λ₯Ό νλ³νλ λ°©λ² : Order embedding
πΉ Subgraph and Motifs
• Building Blocks of Networks → Subgraph (=Mofits) λ λ€νΈμν¬μ function μ΄λ behavior κ³Ό κ°μ μμλ₯Ό κ²°μ νλ€. μ΄λ₯Ό ν΅ν΄ κ·Έλνκ° μ΄λ»κ² μλνλμ§ μ μ μκ³ , κ·Έλν ννμ λ°λΌ μμΈ‘μ μ§νν μ μλ€.
π νΉμ mofits κ° μ 체 κ·Έλνμμ μΌλ§λ λΉλ²νκ² (recurring) λνλλμ§, μΌλ§λ μ€μνμ§μ (significant) μ§μ€
• Graph Isomorphism : λ κ·Έλνκ° μ£Όμ΄μ‘μ λ, identical νμ§ νλ³νλ κ²
• Subgraph Isomorphism : G1 κ³Ό G2 λ ꡬ쑰μ μΌλ‘ λμΌν subgraph λΆλΆμ΄ μ‘΄μ¬νλ€.
• Subgraph Frequency : subgraph μ λΉλλ₯Ό μΈλ λ°©μμ μ κ·Όμλ Graph-level , Node-level, Multiple graphs κ° μλ€.
π Graph - level
π Node - level
π Multiple graphs
• Subgraph Significance : subgraph μ μ€μλλ₯Ό νλ¨νκΈ° μν΄, λλ€νκ² μμ±λ κ·Έλν λ€νΈμν¬μμ mofit λΉλλ³΄λ€ real network μ mofit λΉλκ° λ λ§μ κ²μ΄λΌλ κ°μ μ λμ ν΄ μ€ννλ€.
π Random graph λ₯Ό μμ±νλ λ°©λ² : ER, Configuration model, Switching
π μ€μ κ·Έλνμ λλ€νκ² μμ±λ κ·Έλνμμ κ° motif κ° μΌλ§λ μ€μλλ₯Ό κ°μ§λμ§ Z-score λ₯Ό μ¬μ©ν΄ νκ°νλ€.
πΉ Subgraph Matching
• Query κ·Έλνκ° Target κ·Έλνμ subgraph isomorphism μΈμ§ νμΈνλ Task → Use GNN and embedding space
β node anchor λ₯Ό μ μνμ¬ μλ² λ©μ λΉκ΅
β‘ anchor node μ μ£Όλ³ μ΄μ λ Έλλ€μ μ΄ν΄λ΄
• Order embedding Space : μλ² λ© κ³΅κ°μμ λ κ·Έλν μ€ νλκ° λ€λ₯Έ νλμ μ’νλ¨μ μμΉνλ©΄, μλΈκ·ΈλνλΌ λ³Ό μ μλ€.
π Loss function : max-margin loss
πΉ Finding Frequent subgraphs
• μμ£Ό λ±μ₯νλ mofit μ μ°ΎκΈ° → Representation learning μ ν΅ν΄ μλ² λ© κ³΅κ° λ΄μμ GNN μ νμ©
2οΈβ£ gSpan
• algorithm for Mining frequent subgraphs
• Supports undirected graphs
• DFS Lexicographic order : use lexicographic order (μ¬μ μ μμ) embedding and explore depth-first search (DFS) method
• Using chemical compound datasets, it finds the common chemical compound substructures
'1οΈβ£ AIβ’DS > π GNN' μΉ΄ν κ³ λ¦¬μ λ€λ₯Έ κΈ
[cs224w] Theory of Graph Neural Networks (0) | 2023.01.06 |
---|---|
[CS224W] Graph Neural Network (0) | 2022.11.24 |
[CS224W] Message Passing and Node classification (0) | 2022.11.17 |
[CS224W] PageRank (0) | 2022.11.02 |
[CS224W] 1κ° Machine Learning With Graphs (0) | 2022.10.11 |
λκΈ