Presentation is loading. Please wait.

Presentation is loading. Please wait.

The Matroid Median Problem Viswanath Nagarajan IBM Research Joint with R. Krishnaswamy, A. Kumar, Y. Sabharwal, B. Saha.

Similar presentations


Presentation on theme: "The Matroid Median Problem Viswanath Nagarajan IBM Research Joint with R. Krishnaswamy, A. Kumar, Y. Sabharwal, B. Saha."— Presentation transcript:

1 The Matroid Median Problem Viswanath Nagarajan IBM Research Joint with R. Krishnaswamy, A. Kumar, Y. Sabharwal, B. Saha

2 k-Median Problem Set of locations in a metric space (V,d) Symmetric, triangle inequality Place k facilities such that sum of connection costs (to nearest facility) is minimized: min F µ V, |F| · k  u 2 V d(u,F)

3 k-Median Results poly(log n) approx via tree embeddings [B ’96] LP rounding O(1)-approx [CGST ’99] Lagrangian relaxation + primal dual [JV ’01] Local search with p-exchanges [AGKMMP ’04] best known ratio 3+ ² Hardness of approximation ¼ 1.46 [GK ’98]

4 Red-Blue Median Facilities are of two different types Partition V into red and blue sets Separate bounds k r and k b on facilities Recently introduced [HKK ’10] Motivated by Content Distribution Networks T facility-types (RB Median is T=2) O(1)-approximation ratio via Local Search k r =3 k b =2

5 Matroid Median Given matroid M on ground-set V Locate facilities F that are independent in M Minimize connection cost Recap matroid M=(V, I µ 2 V ) A,B 2 I and |A|<|B| ) 9 e 2 B n A : A [ {e} 2 I Substantial generalization of RB Median The CDN application with T facility-types reduces to partition matroid constraint A B e k 1 =2 k 4 =2 k 2 =3 k 3 =1

6 Talk Outline Thm: 16-approximation for Matroid Median Bad example for Local Search LP relaxation Phase I : sparsification Phase II: reformulation

7 Local Search? Partition matroid with T parts T-1 exchange local search Swap up to T-1 facilities in each step Unlikely to work beyond T=O(1) m m mm m 1 Eg. T=5 Uniform metric on T+1 Clients n=mT+1 OPT = 1 (small fac.) LOPT = m (big fac.) locality gap  (n/T)

8 LP relaxation min  u  v d(u,v) ¢ x uv s.t.  v x uv = 1 8 u 2 V x uv · y v 8 u,v 2 V  v 2 S y v · r(S) 8 S µ V x, y ¸ 0. y 2 M facilities clients u v x uv connection constraints matroid rank constraints

9 Solving the LP Exponential number of rank constraints Use separation oracle: min S µ V r(S) -  v 2 S y v An instance of submodular minimization Also more efficient algorithms to separate over the matroid polytope [C ’84] Solvable in poly-time via Ellipsoid algorithm

10 Idea for approach (1) Problem non-trivial even if metric is a tree  Even O(log n)-approximation not obvious What’s easier than a tree? Suppose input is special star-like instance root facility client 1 client 2 client 3 One root facility (can help any client) Others are private facilities (help only 1 client)

11 Idea for approach (2) Recall LP variables y j : facility opening (in matroid polytope) x ij : connection For any client i, private j 2 P(i) WMA x ij = y j Connection constraint  j x ij = 1 So x ir = 1 -  j 2 P(i) x ij = 1 -  j 2 P(i) y j Can eliminate all connection variables ! r client i private facilities P(i)

12 Idea for approach (3) Reformulate the LP min  i [  j 2 P(i) d ij ¢ y j + d ir ¢ (1-  j 2 P(i) y j ) ] s.t.  j 2 P(i) y j · 1, 8 clients i y 2 M This is just an instance of intersection of M with partition matroid from P(i)s To ensure x ir ¸ 0 matroid constraint x ir x ij

13 Idea for approach (4) Start with LP optimum (x,y) of arbitrary matroid median instance Phase I: Use (x,y) to form clusters of disjoint star-like instances Phase II: Resolve the new star-LP (x,y) itself restricted to the stars not integral Show that new LP is integral ¼ matroid intersection

14 Phase I: sparsify LP solution

15 Outline Modify LP connections x in four steps Similar to [CGST ’99] Key: no change in facility variables y Need to ensure y remains in matroid polytope Not true in [CGST ’99] Require some more (technical) work

16 Step 1: cluster clients L u =  v d uv ¢ x uv, contribution of u to LP obj. B(u) is local ball of u vertices within distance 2 ¢ L u from u Order clients u in increasing L u Pick maximal disjoint set of local balls T are the chosen clients Move each client to T-client close to it 1 2 3 4 5 6 1 2 4 3 5 6 Loss in obj · 4 ¢ LP* (additive)

17 Obs on step 1 Local balls of T clients are disjoint y-value inside any local ball ¸ ½ Markov inequality Restrict to clients T (now weighted) For any p,q 2 T : d(p,q) ¸ 2 ¢ (LP p + LP q ) well separated clients T balls y¸½y¸½ separated

18 More obs on step 1 Suppose y-value in each T’s local ball ¸ 1 Then instance of matroid intersection: Matroid M and partition from local-ball(T) Resolving suitable LP ) integral soln Will need intersection with `laminar’ constraints, not just partition matroid

19 Step 2: private facilities Ensure that each facility in some T-ball or helps at most one client (ie. private) Break connections from all except closest client 1 to facility j Reconnect to facilities in B(1), y-value ¸ ½ Total reconnection for any client · ½ j 1 2 3 Constant factor loss in obj

20 Step 3: uniform objective Each connection from client p to any facility in B(q) will pay same objective d(p,q) Since p,q well separated d(p,q) · O(1) ¢ d(p,j) For any j 2 B(q) Constant factor loss in obj qp

21 Step 4: building stars WMA each client i 2 T connected to Its private facilities P(i), OR Its closest other client k 2 T, ie. facility in B(k) Set of `outer’ connections ¼ directed tree Unique out-edge from each client Lem: Can modify outer connection to `star’ Constant factor loss in obj

22 The star structure One pseudo-root { r, r’ } Every other client connected to either r or r’ All LP-connections x are from client i to: private facility j 2 P(i), obj d(i,j) OR facility in B(k) with k 2 { r, r’ }, uniform obj d(i,k) r r’ i

23 Phase II: using star Will drop all the connection x-variables WMA x ij = y j for j 2 P(i) private facilities Total outer connection=1 -  j 2 P(i) x ij =1 -  j 2 P(i) y j Each outer-connection pays same obj d(i,r) Want property (in integral soln) that P(i)= ; ) there is a recourse connection to r Do not quite ensure this, but…

24 Phase II contd. Add constraint that y(P(r)) + y(P(r’)) ¸ 1 Indeed feasible for (x,y) since each local ball has y-value ¸ ½ This ensures (in integral soln) that P(i)= ; ) there is a recourse connection to r or r’ Lose another constant factor in obj

25 Phase II: new LP Apply constraints for each star to get LP min  i [  j 2 P(i) d ij ¢ y j + d(i,r(i)) ¢ (1-  j 2 P(i) y j ) ] s.t.  j 2 P(i) y j · 1, 8 clients i y(P(r)) + y(P(r’)) ¸ 1, 8 p-root {r, r’} y 2 M Lem: Integral polytope (via proof similar to matroid intersection) matroid constraint laminar constraints

26 Summarize Using LP solution and metric properties reduce to star-like instances Formulate new LP for star-like instances, with only facility variables New LP is integral

27 Other Results O(1)-approximation for prize-collecting version of matroid median Knapsack Median problem (knapsack constraint on open facilities) Give bi-criteria approx, violate budget by w max Can we get true O(1)-approx? Handle other constraints in k-median?

28 Thank You


Download ppt "The Matroid Median Problem Viswanath Nagarajan IBM Research Joint with R. Krishnaswamy, A. Kumar, Y. Sabharwal, B. Saha."

Similar presentations


Ads by Google