approximating node-weighted survivable networks

23
Approximating Node-Weighted Survivable Networks Zeev Nutov The Open University of Israel

Upload: loki

Post on 12-Jan-2016

58 views

Category:

Documents


0 download

DESCRIPTION

Approximating Node-Weighted Survivable Networks. Zeev Nutov The Open University of Israel. Talk Outline. Problem Definition History and Our Results Greedy Algorithm for Node Weighted Steiner Trees Reducing NWSN to Finding Minimum Weight Edge-Cover of Uncrossable Set-Family - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Approximating Node-Weighted  Survivable Networks

Approximating Node-Weighted Survivable Networks

Zeev NutovThe Open University of Israel

Page 2: Approximating Node-Weighted  Survivable Networks

2

Talk Outline

• Problem Definition• History and Our Results• Greedy Algorithm for Node Weighted Steiner Trees• Reducing NWSN to Finding Minimum Weight

Edge-Cover of Uncrossable Set-Family• Spider-Cover Decomposition of Edge-Covers of

Uncrossable Set-Families• Algorithm for Covering Uncrossable Set-Families• Node-Weighted k-Flow is harder than

Densest -Subgraph

Page 3: Approximating Node-Weighted  Survivable Networks

3

Survivable Network (SN)

Instance: A graph G = (V,E), weight function w on edges/nodes, U V, and connectivity requirements r(u,v), u,v U.

Objective: A minimum weight spanning subgraph J of G containing U so that

λJ(u,v) ≥ r(u,v) for all u,v U λJ(u,v) = uv-edge-connectivity in J

Problem Definition

ApproximabilityEdge-weights: 2-approximable [Jain, FOCS 98], APX-hard Node-weights: for r(u,v) {0,1} O(log n)-approximable, Set-Cover hard [Klein and Ravi, IPCO 93]

Page 4: Approximating Node-Weighted  Survivable Networks

4

History and Our ResultsEdge-weights Year Node-weights

2 for r(u,v){0,1} [AKR] 1991

2rmax [WGMV] 1993 2H(n) for r(u,v){0,1} [KR]

2H(rmax) [GGPSTW] 1994

1996 1.35H(n) for r(u,v){0,1}[GK]

2 [J] 1998

Theorem 1NWSN admits a rmax ·3H(n)-approximation algorithm.

Theorem 2ρ-approximation for NWSN with |U|=2 implies 1/ρ2-approximation for Densest -Subgraph.

We do not have a polylogarithmic approximation for any rmax…

But this is not our fault!

What about node-weights and rmax =2?

Page 5: Approximating Node-Weighted  Survivable Networks

5

Node Weighted Steiner TreeInstance: A graph G=(V,E), a set U V of terminals,

weights w(v) for nodes in V−U.Objective: Find a min-weight subtree T of G containing U.

5

a

2 c

31

4

b

d

The “deficiency” of a partial solution I: v(I) = # (components containing terminals in (V,I)) -1.

v(I) = 0w(I) = 5

v(I) = 0w(I) = 8

v(I) = 1w(I) = 7

The “node-weight” w(I) of a partial solution I E: w(I) = w(V(I)) = the weight of endnodes of I

4

3

2

35 3

Page 6: Approximating Node-Weighted  Survivable Networks

6

The Greedy Algorithm

opt

w I

I I S I

Initialize: I While ν(I) > 0 do: Find S E – I so that I I FReturn I.

The Density Condition

optw S

I I S I

Theorem: If ν is decreasing and w is subadditive then the greedy algorithm has approximation ratio ρ ·H(ν()).

Objective:Find in polynomial time an “augmentation” S that satisfies the density condition for “small” ρ.

Page 7: Approximating Node-Weighted  Survivable Networks

7

A Lesson in Zoology

These are also spiders:

In general, a spider is a tree on at least 2 nodes, which has at most one node of degree ≥ 3.

This is a spider:

Page 8: Approximating Node-Weighted  Survivable Networks

8

Spider Decomposition of Trees

Center – The single node of degree ≥ 3. If there is no node of degree ≥ 3, any node can be a center.Leaves – The non-center nodes of degree 1.

Lemma: Every tree can be decomposed into node-disjoint spiders such that every leaf of the tree belongs to a unique spider.

1. Select a node v whose sub-tree is a spider.

2. Remove v and its sub-tree.3. Remove the path from v to its

closet ancestor of degree 3.4. Repeat.

Page 9: Approximating Node-Weighted  Survivable Networks

9

Finding the First Augmentation

Finding a spider S (in fact, a Shortest Path Tree) of optimal density: For each node s in the graph

1. Sort the paths from s to terminals in increasing weight order.2. Add the two lightest paths.3. Add paths in increasing weight order, till reaching minimum density.

23 74

Terminals: 2Weight: 8Density: 8 = 8/(2-1)

Terminals: 4Weight: 19Density: ~ 6.3=19/(4-1)

T = optimal tree; we may assume: terminals = leaves of T

iw T w S leaves / 2i iS S

By averaging, there is a spider Si such that:

2i

i

w S w T

S T

2 iT S

Terminals: 3Weight: 12Density: 6 = 12/(3-1)

{Si} – spider-decomposition of T. The spiders are disjoint

3

Page 10: Approximating Node-Weighted  Survivable Networks

10

The Complete Algorithm

Finding an augmentation with a general partial cover I:1. Contract every connected component of (V,I) into a

super-node; a super node is a super-terminal if it contains a terminal.

2. Find an augmentation in the new graph (the partial cover is now ).

The previous algorithm finds an augmentation obeying the Density Condition with ρ=2 if the current partial cover is I = .

The approximation ratio of the algorithm is 2H(|U|).

Page 11: Approximating Node-Weighted  Survivable Networks

11

Algorithm for NWSN

The algorithm has rmax iterations.In iteration k we find a 3H(n)-approximation for the problem:

Given: A graph J=Jk-1 with λJ(u,v) ≥ min{r(u,v),k-1} for all u,v UFind: An edge set I with w(V(I)) minimum so that λJ+I(u,v) ≥ min{r(u,v),k} for all u,v U

Hence after rmax iterations, a feasible solution of weight at most rmax ·3H(n)·opt is found.

Instance: A graph G = (V,E), weight function w on the nodes, U V, and connectivity requirements r(u,v), u,v U.

Objective: A minimum weight spanning subgraph J of G containing U so that λJ(u,v) ≥ r(u,v) for all u,v U.

Page 12: Approximating Node-Weighted  Survivable Networks

12

Covers of Uncrossable Set-Families

Node-Weighted Set-Family Edge-Cover (NWSFC)

The augmentation problem we want to solve is a particular case of the following problem:

Instance: A graph (V,E), node weights {w(v):v V}, and an uncrossable set-family on V.Objective: Find an -cover I ⊆ E of minimum node-weight (edge e covers set X if e has exactly one endnode in X)

is uncrossable if X,Y implies at least one of the following:

orX ∩Y, X Y

X − Y, Y −X X

Y V−Y

V−XNote: The inclusion minimal members of are pairwise disjoint.

Page 13: Approximating Node-Weighted  Survivable Networks

13

Spider-Covers of Uncrossable Set-Families• () = the family of inclusion minimal sets in (min-cores)• (C) = sets in that contain a unique min-core C (cores) • (s,C) = {X (C) : s V-X}• (s,) = {(s,C) : C }

Definition: Let ⊆ () and let sV. An edge set S is an (s,)-cover if: - S covers (s,C) for every C- if ={C} then no member of (C) contains sAn (s,)-cover S is a spider-cover if it can be partitioned into (s,C)-covers {SC:C } so that the node sets {V(SC) −s} are pairwise disjoint.

ss

Page 14: Approximating Node-Weighted  Survivable Networks

14

Spider-Coves Decompositions

Definition of a Spider-Cover Decomposition:A sub-partition S1,…,Sq of a cover I is a spider coverdecomposition of I if there exists a partition 1, …,q of () and centers s1,…,sq V so that:- Each Si is an (si,i)-cover - The node sets V(Si) are pairwise disjoint.

Spider-Cover Decomposition Theorem:Any uncrossable family cover has a spider-cover decomposition.

Proof: Later.

Page 15: Approximating Node-Weighted  Survivable Networks

15

Covering Uncrossable Families

S is a spider with leaves: Δ(S) ≥ /2 (tight for =2)S is an (s,)-cover with ||=: Δ(S) ≥ /3 (tight for =3)

= |()| = # (min-cores)Δ(S) = decrease in the deficiency caused by adding S to the partial solution

optw S

S

Density Condition (for I=)

If S is an (s,)-cover then Δ(S) ≥ (||-1)/2 if || ≥2 Δ(S) = 1 if || =1

The Spider-Cover Lemma

Tight Example

u0

v1

v2

u2

u1

s

Page 16: Approximating Node-Weighted  Survivable Networks

16

The Algorithm

Thus the Greedy Algorithm can be implemented in polynomial time with ρ=3.

Approximation ratio: 3H(()) = 3H(|()|) ≤ 3H(n)

The Spider-Cover Lemma implies that there exists a spider-cover that satisfies the Density Condition with ρ=3.

Such spider-cover can be found in polynomial time assuming we can compute in polynomial time:

- The family () of min-cores (max-flows)- Minimum weight (s,C)-cover (min-cost k-flows)

Page 17: Approximating Node-Weighted  Survivable Networks

17

The Spider-Cover Decomposition Thm – Proof Sketch

We may assume that I is an inclusion minimal -cover.Then for every eI there exists a witness set We , namely:

e is the unique edge in I that covers We.

A family = {We : e I} is called a witness family for I

(every eI has a unique witness set in We ).

Notation – uncrossable family (X,Y implies X ∩Y, X Y or X−Y,Y−X)I – an -cover (for any X there is eI with exactly one endnode in X)

Lemma:Let I be an inclusion minimal cover of an uncrossable family . Then there exists a witness family for I which is laminar.

Page 18: Approximating Node-Weighted  Survivable Networks

18

The Spider-Cover Decomposition Thm – Proof Sketch

For a min-core C() define:• LC = the maximal set in containing C• eC = the unique edge in I covering LC, eC=sCvC, vC LC

• SC = edges in I contained in LC plus eC

Assumptions:– Every member of is a core – Every minimal member (leaf) of is a min-core.

CL

eC C

C

C

s

v

Page 19: Approximating Node-Weighted  Survivable Networks

19

The Spider-Cover Decomposition Thm – Proof Sketch

Lemma:• The sets {LC : C in ()} are pairwise disjoint.• The sets {SC : C in ()} are pairwise disjoint.• SC covers all cores contained in LC. Corollary:Any partition 1, …, q of () induces a partition S1,…,Sq of I. We seek a partition so that S1,…,Sq is a spider-cover decomposition.

- A natural partition of () is by the stars of {eC : C()}.

- This approach fails for 1-edge stars; SC is not a spider-cover if there is a dangerous set MC containing LC+sC

- Every star with at least 2 edges indeed induces a spider-cover.

CL

eC

CL

MC

eC'eC

C

C

C

s

v

Page 20: Approximating Node-Weighted  Survivable Networks

20

The Spider-Cover Decomposition Thm – Proof Sketch

CL

MCeCeC'

s

How do we group dangerous cores? - group some together, or - assign to “non-dangerous” stars.

Assigning singleton classes:

Every singleton class {MC} of is assigned to the part of any edge eC’ covering MC .

Observation: Every dangerous MC is covered by some edge eC’ .

CL

MCeCeC'

s

Grouping dangerous cores together:

The relation ={(C,C’) : MC ∩ MC’ ≠ } is an equivalence, and its classes of size ≥2 induce spider-covers(center − any node in the intersection of MC’s)

Page 21: Approximating Node-Weighted  Survivable Networks

21

min : , .X X A B I X k

-appr f or NWSN

21 2

max : ,I X X A B X

-appr f or D -S

min : , , 1..

.

min ,

k

k

X X A B I X k k I

k X

X X k

X A B I X k

1. Run the -approximation algorithm f or

f or all .

For every we have a (possibly empty) set

2. Set where is the largest integer so that

and .

2 2

2 2 2

opt .

'

1.

22 2

I X D S

X X X

I X I X I X I XX

Note that

3. Find with so that

B

A

s

t

( , ) ( , ) 0

, 1,

, : :

k r s t k r u v

J A B I w v v A B

s t sa a A bt b B

Node-Weighted k-Flow (NW F): and otherwise.

Given an instance , of bipartite D S, set

add nodes and edges of capacity each.

I

Reducing NWSN to bipartite DS

Page 22: Approximating Node-Weighted  Survivable Networks

22

Summary and Open QuestionsWhat did we do?Generalized the decomposition of a tree into spiders to covers of uncrossable families (looks easy after found…)What do we get?E.g., an rmax·3H(n)-approximation algorithm for NWSN.Any other applications?Probably YES.

Open Question:Node-Weighted k-Flow (NWkF) is a special case of

NWSN where r(s,t)=k and r(u,v)=0 otherwise. NWkF admits a k-approximation algorithm.Anything better, even for unit weights?

Page 23: Approximating Node-Weighted  Survivable Networks