SLIDE 1 Local and Union Page Numbers
Laura Merker
Karlsruhe Institute of Technology
Torsten Ueckerdt∗
Karlsruhe Institute of Technology
Graph Drawing 2019 September 20, 2019 Pruhonice
SLIDE 2
book embedding (≺, P) ⊲ linear vertex ordering ≺ ⊲ edge partition P = {P1, . . . , Pk} ⊲ u ≺ x ≺ v ≺ y, uv ∈ Pi, xy ∈ Pj ⇒ i = j spine ordering pages each page crossing-free K5 K3,3 u v x y
SLIDE 3
book embedding (≺, P) ⊲ linear vertex ordering ≺ ⊲ edge partition P = {P1, . . . , Pk} ⊲ u ≺ x ≺ v ≺ y, uv ∈ Pi, xy ∈ Pj ⇒ i = j spine ordering pages each page crossing-free K5 K3,3 u v x y k-local book embedding: each vertex on at most k pages
SLIDE 4
book embedding (≺, P) ⊲ linear vertex ordering ≺ ⊲ edge partition P = {P1, . . . , Pk} ⊲ u ≺ x ≺ v ≺ y, uv ∈ Pi, xy ∈ Pj ⇒ i = j spine ordering pages each page crossing-free K5 K3,3 u v x y k-union embedding: each page crossing-free components
SLIDE 5
page number pn(G) = min k: ∃ k-page book embedding K3,3 K5 pnℓ pnu pn 2 2 3 2 3 3 each page crossing-free minimize # pages union page number pnu(G) = min k: ∃ k-union embedding local page number pnℓ(G) = min k: ∃ k-local book embedding minimize # pages each page union of crossing-free components each page crossing-free minimize # pages at any one vertex
SLIDE 6
Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G).
SLIDE 7 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . |E| =
#{edges in P}
SLIDE 8 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . |E| =
#{edges in P} <
2 · #{vertices on P} as each page is outerplanar
SLIDE 9 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . |E| =
#{edges in P} <
2 · #{vertices on P} ≤ 2 · pnℓ(G)|V | as each page is outerplanar as each vertex is on at most pnℓ(G) pages
SLIDE 10 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . |E| =
#{edges in P} <
2 · #{vertices on P} ≤ 2 · pnℓ(G)|V | Hence pnℓ(G) ≥ |E| 2|V | = 1 4 · avd(G) as each page is outerplanar as each vertex is on at most pnℓ(G) pages avd(G) =
|V | = 2|E| |V | as each vertex is on at most pnℓ(G) pages
SLIDE 11 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . |E| =
#{edges in P} pnℓ(G) ≥ 1 4 mad(G) <
2 · #{vertices on P} ≤ 2 · pnℓ(G)|V | Hence pnℓ(G) ≥ |E| 2|V | = 1 4 · avd(G) = ⇒ as each page is outerplanar as each vertex is on at most pnℓ(G) pages avd(G) =
|V | = 2|E| |V | as each vertex is on at most pnℓ(G) pages
SLIDE 12
Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). . . . gives also an upper bound mad(G) = k A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G)
SLIDE 13 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). . . . gives also an upper bound mad(G) = k = ⇒
- rientation with
- utdeg(v) ≤ k/2 + 1
A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G)
SLIDE 14 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). . . . gives also an upper bound mad(G) = k = ⇒
- rientation with
- utdeg(v) ≤ k/2 + 1
= ⇒ (k/2 + 2)-local star partition A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G)
SLIDE 15 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). . . . gives also an upper bound mad(G) = k = ⇒
- rientation with
- utdeg(v) ≤ k/2 + 1
= ⇒ (k/2 + 2)-local star partition = ⇒ pnℓ(G) ≤ 1 2 mad(G) + 2 A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G) as stars are crossing-free
SLIDE 16
Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G) . . . gives also an upper bound pnℓ(G) ≤ 1 2 mad(G) + 2
SLIDE 17
Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G) . . . gives also an upper bound pnℓ(G) ≤ 1 2 mad(G) + 2 mad(G) = k = ⇒ k + 2 star forests partition = ⇒ pnu(G) ≤ mad(G) + 2 . . . also for union page number
SLIDE 18 Comparison of variants ⊲ For any graph G we have pnℓ(G) ≤ pnu(G) ≤ pn(G). A simple lower bound . . . pnℓ(G) ≥ 1 4 mad(G) . . . gives also an upper bound pnℓ(G) ≤ 1 2 mad(G) + 2 mad(G) = k = ⇒ k + 2 star forests partition = ⇒ pnu(G) ≤ mad(G) + 2 . . . also for union page number Corollary. pnu(G) ≤ 4 pnℓ(G) + 2 but there are n-vertex k-regular graphs with pnu(G) ≤ k + 2 and pn(G) = Ω √ kn
1 2 − 1 k
- “local and union page numbers
are tied to density, classical page number is tied to structure”
SLIDE 19 Planar graphs pn(G) = 3
planar graphs pnℓ pnu pn 3 4 2
1 4 mad(G) ≤ pnℓ(G) ≤ pnu(G)
non-hamiltonian triangulation
SLIDE 20 Planar graphs pn(G) = 3 pnℓ(G) = pnu(G) = 2
planar graphs pnℓ pnu pn 3 4 2
SLIDE 21 Planar graphs pn(G) = 3 pnℓ(G) = pnu(G) = 2
planar graphs pnℓ pnu pn 3 4 2 Theorem. pnu(G) ≥ pnℓ(G) ≥ 3 there is a planar graph G with
✗ ✗
SLIDE 22 Planar graphs pn(G) = 3 pnℓ(G) = pnu(G) = 2
planar graphs pnℓ pnu pn 3 4 2
✗ ✗
G planar = ⇒ = ⇒ pnℓ(G) ≤ 4
- rientation with
- utdeg(v) ≤ 3
= ⇒ 4-local star partition = ⇒ 5 star forest partition = ⇒ pnu(G) ≤ 5 G planar
SLIDE 23 k-Trees
(graphs of treewidth k)
1-tree:
k-tree:
K1 attach to K1 Kk attach to Kk
SLIDE 24 k-Trees
(graphs of treewidth k)
1-tree:
k-tree:
K1 attach to K1 Kk attach to Kk
k-trees pnℓ pnu pn k k + 1 k/2
1 4 mad(G) ≤ pnℓ(G) ≤ pnu(G)
|E| ≈ k|V | = ⇒ mad(G) ≈ 2k k − 1 . . .
SLIDE 25 k-Trees
(graphs of treewidth k)
1-tree:
k-tree:
K1 attach to K1 Kk attach to Kk
k-trees pnℓ pnu pn k k + 1 k/2 Theorem. ℓ-local book embedding for every k-tree ℓ-local book embedding for every k-tree with a forest on each page = ⇒
✗ ✗
k − 1
✗ ✗
. . . . . . . . .
SLIDE 26 k-Trees
(graphs of treewidth k)
1-tree:
k-tree:
K1 attach to K1 Kk attach to Kk
k-trees pnℓ pnu pn k k + 1 k/2
✗ ✗
k − 1
✗ ✗
. . . . . . . . . G k-tree = ⇒ = ⇒ pnℓ(G) ≤ k + 1
- rientation with
- utdeg(v) ≤ k
= ⇒ (k + 1)-local star partition = ⇒ k + 1 star forest partition = ⇒ pnu(G) ≤ k + 1 G k-tree
SLIDE 27 k-Trees
(graphs of treewidth k)
1-tree:
k-tree:
K1 attach to K1 Kk attach to Kk
k-trees pnℓ pnu pn k k + 1 k/2
✗ ✗
k − 1
✗ ✗
A possible approach? ⊲ consider the unique (k + 1)-coloring of G ⊲ then any two color classes induce a tree k trees at each vertex can be combined to k or k + 1 forests Still open: Find the spine ordering! . . . . . . . . .
SLIDE 28 Complete graphs pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . . . .
n 3
Kn pnℓ(K6) = 2
SLIDE 29 Complete graphs pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . . . .
n 3
Kn pnℓ(K6) = 2
SLIDE 30 Complete graphs pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . . . .
n 3
Kn pnℓ(K6) = 2
SLIDE 31 Complete graphs pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . . . .
n 3
Kn pnℓ(K6) = 2 pnℓ(K9) = 3
SLIDE 32 Complete graphs pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . . . .
n 3
Kn pnℓ(K6) = 2 pnℓ(K9) = 3 pnℓ(K11) = 4
SLIDE 33 Complete graphs pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . . . .
n 3
Kn pnℓ(K6) = 2 pnℓ(K9) = 3 pnℓ(K11) = 4 pnℓ(K15) ≤ 5
SLIDE 34 pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . complete graphs, Kn Open problems pnℓ pnu pn k k + 1 pnℓ pnu pn 3 4 ⊲ computational complexity ? ⊲ maximum pnu(G)/ pnℓ(G) ? ⊲ Km,n ? ⊲ local and union queue numbers ? planar graphs k-trees, treewidth k
SLIDE 35 pnℓ pnu pn ⌈ n
2 ⌉
⌈ n−1
4 ⌉
. . . complete graphs, Kn Open problems pnℓ pnu pn k k + 1 pnℓ pnu pn 3 4 ⊲ computational complexity ? ⊲ maximum pnu(G)/ pnℓ(G) ? ⊲ Km,n ? ⊲ local and union queue numbers ? planar graphs k-trees, treewidth k
Thank you