Improving Coarsening Schemes for Hypergraph Partitioning by Exploiting Community Structure
SEA’17 · June 23, 2017 Tobias Heuer and Sebastian Schlag
INSTITUTEOF THEORETICAL INFORMATICS · ALGORITHMICS GROUP
KIT – University of the State of Baden-Wuerttemberg and National Laboratory of the Helmholtz Association www.kit.edu Hypergraphs
Generalization of graphs ⇒ hyperedges connect ≥ 2 nodes
Graphs ⇒ dyadic (2-ary) relationships Hypergraphs ⇒ (d-ary) relationships
Hypergraph H = (V , E, c, ω) e1 e4 v2 v Vertex set V = {1, ..., n} v 12 11 v3
Edge set E ⊆ P (V ) \ ∅ v10 v1 v13 Node weights c : V → R≥1 v8 v7 e v 5 v e 9 4 Edge weights ω : E → R≥1 3
v6 v5 e2
1 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Hypergraphs
Generalization of graphs ⇒ hyperedges connect ≥ 2 nodes
Graphs ⇒ dyadic (2-ary) relationships Hypergraphs ⇒ (d-ary) relationships pin net Hypergraph H = (V , E, c, ω) e1 e4 v2 v Vertex set V = {1, ..., n} v 12 11 v3
Edge set E ⊆ P (V ) \ ∅ v10 v1 v13 Node weights c : V → R≥1 v8 v7 e v 5 v e 9 4 Edge weights ω : E → R≥1 3
v P P v6 5 e2 |P| = e∈E |e| = v∈V d(v)
1 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Hypergraph Partitioning Problem
Partition hypergraph H = (V , E, c, ω) into k disjoint blocks
Π = {V1, ... , Vk } such that:
blocks Vi are roughly equal-sized: c(V ) c(Vi ) ≤ (1 + ε) k connectivity objective is minimized: V1 V2
V 4 V3
2 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Hypergraph Partitioning Problem
Partition hypergraph H = (V , E, c, ω) into k disjoint blocks
Π = {V1, ... , Vk } such that: imbalance blocks Vi are roughly equal-sized: parameter c(V ) c(Vi ) ≤ (1 + ε) k connectivity objective is minimized: V1 V2
V 4 V3
2 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Hypergraph Partitioning Problem
Partition hypergraph H = (V , E, c, ω) into k disjoint blocks
Π = {V1, ... , Vk } such that: imbalance blocks Vi are roughly equal-sized: parameter c(V ) c(Vi ) ≤ (1 + ε) k connectivity objective is minimized: V1 V2
V 4 V3
2 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Hypergraph Partitioning Problem
Partition hypergraph H = (V , E, c, ω) into k disjoint blocks
Π = {V1, ... , Vk } such that: imbalance blocks Vi are roughly equal-sized: parameter c(V ) c(Vi ) ≤ (1 + ε) k connectivity objective is minimized: P V1 V2 e∈cut(λ−1) ω(e)=6
connectivity: # blocks connected by net e V 4 V3
2 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Hypergraph Partitioning Problem
Partition hypergraph H = (V , E, c, ω) into k disjoint blocks
Π = {V1, ... , Vk } such that: imbalance blocks Vi are roughly equal-sized: parameter c(V ) c(Vi ) ≤ (1 + ε) k connectivity objective is minimized: P V1 V2 e∈cut(λ−1) ω(e)=6
connectivity: # blocks connected by net e V 4 V3
2 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Applications
VLSI Design Scientific Computing 0 1 2 3 4 5 6 7 0 ××× 1 × × 2 ×××× Application 3 ××× 4 ××× Domain 5 ×× 6 ×××××××× 7 ××
Hypergraph Model
facilitate Goal minimize floorplanning & placement communication
3 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group The Multilevel Framework
input hypergraph output partition
··· ··· match / cluster local search
contract uncontract
··· ··· initial partitioning
4 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group This Talk: Coarsening Phase
input hypergraph output partition
··· ··· match / cluster local search
contract uncontract
··· ··· initial partitioning
5 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Common Strategy: avoid global decisions local, greedy algorithms
Objective: identify highly connected vertices using...
1 foreach vertex v do 2 cluster[v] := argmax rating(v,u) neighbor u
6 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Common Strategy: avoid global decisions local, greedy algorithms
Objective: identify highly connected vertices using...
1 foreach vertex v do 2 cluster[v] := argmax rating(v,u) neighbor u
Main Design Goals:[Karypis,Kumar 99] 1: reduce size of nets easier local search 2: reduce number of nets easier initial partitioning 3: maintain structural similarity good coarse solutions
6 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search 2: reduce number of nets easier initial partitioning
7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search 2: reduce number of nets easier initial partitioning foo ⇒hypergraph-tailored rating functions: P ω(e) r(u, v) := |e|−1 net e containing u,v
7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search 2: reduce number of nets easier initial partitioning foo ⇒hypergraph-tailored rating functions: P ω(e) r(u, v) := |e|−1 net e large number ... containing u,v
7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search 2: reduce number of nets easier initial partitioning foo ⇒hypergraph-tailored rating functions: P ω(e) of heavy nets ... r(u, v) := |e|−1 net e large number ... containing u,v
7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search 2: reduce number of nets easier initial partitioning foo ⇒hypergraph-tailored rating functions: P ω(e) of heavy nets ... r(u, v) := |e|−1 net e ... with small size large number ... containing u,v
7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search X 2: reduce number of nets easier initial partitioning X foo ⇒hypergraph-tailored rating functions: P ω(e) of heavy nets ... r(u, v) := |e|−1 net e ... with small size large number ... containing u,v
7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search X 2: reduce number of nets easier initial partitioning X foo ⇒hypergraph-tailored rating functions: P ω(e) of heavy nets ... r(u, v) := |e|−1 net e ... with small size foolarge number ... containing u,v 3:foomaintain structural similarity good coarse solutions ⇒prefer clustering over matching ⇒ensure ∼balanced vertex weights 7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Clustering-based Coarsening
Main Design Goals: 1: reduce size of nets easier local search X 2: reduce number of nets easier initial partitioning X foo ⇒hypergraph-tailored rating functions: P ω(e) of heavy nets ... r(u, v) := |e|−1 net e ... with small size foolarge number ... containing u,v 3:foomaintain structural similarity good coarse solutions ⇒prefer clustering over matching ensure ∼balanced vertex weights ⇒ enough? 7 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group What could possibly go wrong? ... a lot:
n1 u v {u, v} input obscured clusters
8 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group What could possibly go wrong? ... a lot:
n1 u v {u, v} input obscured clusters
maximal matching random tie-breaking ISSUES
prefer unclustered heavy neighbors
8 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group What could possibly go wrong? ... a lot:
n1 u v {u, v} input obscured clusters
maximal matching random tie-breaking ISSUES
prefer unclusteredfoo heavy neighbors ⇒Problem: relying only on local information!
8 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Our Approach: Community-aware Coarsening
n1 u v {u, v} input obscured clusters
9 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Our Approach: Community-aware Coarsening
n1 u v {u, v} input obscured clusters
SOLUTION community detection
9 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Our Approach: Community-aware Coarsening
n1 u v {u, v} input obscured clusters
SOLUTION community detection
Framework: preprocessing: determine community structure only allow intra-community contractions
9 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Our Approach: Community-aware Coarsening
n1 u v {u, v} input obscured clusters
SOLUTION community detection Framework: How? preprocessing: determine community structure only allow intra-community contractions
9 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Detecting Community Structure
Goal: partition graph into natural groups C Community: internally dense, externally sparse subgraph
10 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Detecting Community Structure
Goal: partition graph into natural groups C Community: internally dense, externally sparse subgraph
(One) Formalization: [Newman,Girvan 04] fraction of Modularity mod(G, C) := cov(G, C) − E[cov(G, C)] intra-cluster edges P |E(C)| Coverage cov(G, C) := |E| C∈C
10 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Detecting Community Structure
Goal: partition graph into natural groups C Community: internally dense, externally sparse subgraph
(One) Formalization: [Newman,Girvan 04] fraction of Modularity mod(G, C) := cov(G, C) − E[cov(G, C)] intra-cluster edges P |E(C)| Coverage cov(G, C) := |E| C∈C
Efficient Heuristic: Louvain Method (multilevel, local, greedy) [Blondel et al. 08] repeatedly move nodes to neighbor communities coarsen graph & repeat
10 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Graph Representations of Hypergraphs Hypergraph
11 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Graph Representations of Hypergraphs Hypergraph
Clique Graph
destroys natural sparsity: |e| edges per net e - 2 -exaggerates importance of large nets
11 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Graph Representations of Hypergraphs Hypergraph
Clique Graph
destroys natural sparsity: |e| edges per net e - 2 -exaggerates importance of large nets
Bipartite (Star) Graph +compact representation: O(|P|) space -nets become nodes & part of clustering
11 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
Density: d := m = |P|/n = d(v) n |P|/m |e|
12 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
Density: d := m = |P|/n = d(v) n |P|/m |e|
d ≈ 1
|V | ' |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) d(v) ' |e|
12 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
Density: d := m = |P|/n = d(v) n |P|/m |e|
d ≈ 1 d 1
|V | ' |E| |V | |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) d(v) ' |e| d(v) |e|
12 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
Density: d := m = |P|/n = d(v) n |P|/m |e|
d 1 d ≈ 1 d 1
|V | |E| |V | ' |E| |V | |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) d(v) |e| d(v) ' |e| d(v) |e|
12 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
d 1 d ≈ 1 d 1 |V | |E| |V | ' |E| |V | |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) foo d(v) |e| d(v) ' |e| d(v) |e| ⇒addressed via edge weights:
13 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
d 1 d ≈ 1 d 1 |V | |E| |V | ' |E| |V | |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) foo d(v) |e| d(v) ' |e| d(v) |e| ⇒addressed via edge weights: ω(v, e) := 1 baseline
13 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
d 1 d ≈ 1 d 1 |V | |E| |V | ' |E| |V | |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) foo d(v) |e| d(v) ' |e| d(v) |e| ⇒addressed via edge weights: ω(v, e) := 1 baseline
1 ωe(v, e) := |e| small nets higher influence
13 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Bipartite Graphs: Modeling Pecularities
d 1 d ≈ 1 d 1 |V | |E| |V | ' |E| |V | |E| Hypernodes (>-nodes) Hyperedges (⊥-nodes) foo d(v) |e| d(v) ' |e| d(v) |e| ⇒addressed via edge weights: ω(v, e) := 1 baseline
1 ωe(v, e) := |e| small nets higher influence
d(v) + high degree higher influence ωde(v, e) := |e|
13 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experiments – Benchmark Setup
System: 1 core of 2 Intel Xeon E5-2670 @ 2.6 Ghz, 64 GB RAM
Hypergraphs:1
Application VLSI Sparse Matrix SAT Solving Benchmark Set ISPD98 DAC2012 UF-SPM SAT14 Representation direct direct row-net literal primal dual Density Class d ≈ 1 d ≈ 1 d 1, d ≈ 1, d 1 d 1 d 1 d 1 Community Str. XX some instances XXX # Hypergraphs 18 10 184 92 92 92 # in Subset 10 5∗ 60 30 30∗ 30∗
k ∈ {2, 4, 8, 16, 32, 64, 128} with imbalance: ε = 3% 8 hours time limit / instance
Comparing KaHyPar-CA with: KaHyPar-K hMetis-R & hMetis-K PaToH-Default & PaToH-Quality 1available @ https://algo2.iti.kit.edu/schlag/sea2017/
14 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%]
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%]
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%] d(v) ωde(v, e) := 15 |e| 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%] d(v) ωde(v, e) := 15 |e| 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%] d(v) ωde(v, e) := ω(v, e) := 1 15 |e| 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%] d(v) ωde(v, e) := ω(v, e) := 1 15 |e| 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Comparison of Edge Weighting Schemes
d 1 d 1 d 1 ≪ ≈ ≫ 75 20 75 50 25 10 50 0 0 25 -25 -10 0 -50 -20 -75 -25 Initial Cut Improv. [%] d(v) ωde(v, e) := ω(v, e) := 1 ω(v, e) := 1 15 |e| 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Avg. Cut Improv.-15 [%] -5.0 -15
15 5.0 15 10 10 2.5 5 5 0 0.0 0 -5 -5 -2.5 -10 -10
Min Cut Improv.-15 [%] -5.0 -15 CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde) CA(ω) CA(ωe)CA(ωde)
15 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results – Partitioning Quality
Example 1.00 0.80 0.60 0.40
0.20 Best 0.10 Algorithm 0.05
− Algorithm 1 0.01 Algorithm 2 1 Algorithm 3
0.00
1 100 200 300 400 600 800 10001200 1600 2000 # Instances
16 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results – Partitioning Quality
Example 1.00 0.80 0.60 0.40
0.20 Best 0.10 Algorithm 0.05
− Algorithm 1 0.01 Algorithm 2 1 Algorithm 3
0.00
1 100 200 300 400 600 800 10001200 1600 2000 # Instances
16 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results – Partitioning Quality
Example 1.00 0.80 0.60 0.40
0.20 Best 0.10 Algorithm 0.05
− Algorithm 1 0.01 Algorithm 2 1 Algorithm 3
0.00
1 100 200 300 400 600 800 10001200 1600 2000 # Instances
16 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results – Partitioning Quality
Example 1.00 0.80 0.60 0.40
0.20 Best 0.10 Algorithm 0.05
− Algorithm 1 0.01 Algorithm 2 1 Algorithm 3
0.00
1 100 200 300 400 600 800 10001200 1600 2000 # Instances
16 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results – Partitioning Quality
Example 1.00 0.80 0.60 0.40
0.20 Best 0.10 Algorithm 0.05
− Algorithm 1 0.01 Algorithm 2 1 Algorithm 3
0.00
1 100 200 300 400 600 800 10001200 1600 2000 # Instances
16 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results - Quality
d ≈ 1 :
d ≈ 1 : d 1 :
17 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results - Quality
d 1 : d 1 :
18 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results - Running Time
Running Time [s] Algorithm All DAC2012 ISPD98 Primal Literal Dual SPM WebSocial KaHyPar 20.4 289.5 8.1 15.6 30.6 57.8 10.9 66.7 KaHyPar-CA 31.0 369.0 12.3 32.9 64.7 68.3 13.9 67.1 hMetis-R 79.2 446.4 29.0 66.2 142.1 200.4 41.8 89.7 hMetis-K 57.9 240.9 23.2 44.2 94.9 125.6 36.0 111.9 PaToH-Q 5.9 28.3 1.9 6.9 9.2 10.6 3.4 4.7 PaToH-D 1.2 6.5 0.4 1.1 1.6 2.9 0.8 0.9
19 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Experimental Results - Running Time
Running Time [s] Algorithm All DAC2012 ISPD98 Primal Literal Dual SPM WebSocial KaHyPar 20.4 289.5 8.1 15.6 30.6 57.8 10.9 66.7 KaHyPar-CA 31.0 369.0 12.3 32.9 64.7 68.3 13.9 67.1 hMetis-R 79.2 446.4 29.0 66.2 142.1 200.4 41.8 89.7 hMetis-K 57.9 240.9 23.2 44.2 94.9 125.6 36.0 111.9 PaToH-Q 5.9 28.3 1.9 6.9 9.2 10.6 3.4 4.7 PaToH-D 1.2 6.5 0.4 1.1 1.6 2.9 0.8 0.9
19 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Conclusion & Discussion
KaHyPar-CA - Community-aware Coarsening
Community Detection via: modularity maximization Louvain Method (LM) bipartite, weighted graph
Future Work: speedup preprocessing: parallel LM resolution limit multi-resolution modularity other formalizations: KaHyPar-Framework Infomap Open-Source on Github: https://git.io/vMBaR Surprise
20 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group References
[Newman, Girvan 04] M. E. J. Newman and M. Girvan. Finding and evaluating community structure in networks. Physical Review E, 69:026113, Feb 2004. [Blondel et al. 08] V. D. Blondel, J. Guillaume, R. Lambiotte, and E. Lefebvre. Fast unfolding of com- munities in large networks. Journal of Statistical Mechanics: Theory and Experi- ment, 2008(10):P10008, 2008. [Karypis, Kumar 99] G. Karypis and V. Kumar. Multilevel K-way Hypergraph Partitioning. In Proceed- ings of the 36th ACM/IEEE Design Automation Conference, pages 343–348. ACM, 1999.
21 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group Taxonomy of Hypergraph Partitioning Tools
Recursive Bisection Direct k-way MLPart 1998 PaToH hMetis-R VLSI hMetis-K 1999 Sparse Matrices Mondriaan 2005 Zoltan parallel 2006 Parkway kPaToH 2008 multi-constraint UMPa multi-objective 2013
KaHyPar-R n-Level 2016
KaHyPar-K n-Level 2017
22 Sebastian Schlag – Community-aware Coarsening Schemes for Hypergraph Partitioning Institute of Theoretical Informatics Algorithmics Group