## Algorithms for Facility Location Problems with Outliers (Extended Abstract) (2000)

### Cached

### Download Links

- [www.msci.memphis.edu]
- [www.cs.umd.edu]
- [ce.sharif.ac.ir]
- DBLP

### Other Repositories/Bibliography

Venue: | In Proceedings of the 12th Annual ACM-SIAM Symposium on Discrete Algorithms |

Citations: | 70 - 8 self |

### BibTeX

@INPROCEEDINGS{Charikar00algorithmsfor,

author = {Moses Charikar and Samir Khuller and David M. Mount and Giri Narasimhan},

title = {Algorithms for Facility Location Problems with Outliers (Extended Abstract)},

booktitle = {In Proceedings of the 12th Annual ACM-SIAM Symposium on Discrete Algorithms},

year = {2000},

pages = {642--651}

}

### Years of Citing Articles

### OpenURL

### Abstract

) Moses Charikar Samir Khuller y David M. Mount z Giri Narasimhan x Abstract Facility location problems are traditionally investigated with the assumption that all the clients are to be provided service. A significant shortcoming of this formulation is that a few very distant clients, called outliers, can exert a disproportionately strong influence over the final solution. In this paper we explore a generalization of various facility location problems (K-center, K-median, uncapacitated facility location etc) to the case when only a specified fraction of the customers are to be served. What makes the problems harder is that we have to also select the subset that should get service. We provide generalizations of various approximation algorithms to deal with this added constraint. 1 Introduction The facility location problem and the related clustering problems, k-median and k-center, are widely studied in operations research and computer science [3, 7, 22, 24, 32]. Typically in...

### Citations

2334 | Algorithms for Clustering Data - Jain, Dubes - 1988 |

1308 |
Robust Regression and Outlier Detection
- Rousseeuw, Leroy
- 1987
(Show Context)
Citation Context ...ions can be applied to any ofsthe variants of the facility location problems. The ter-sminology used in the first formulation is borrowed fromsthe well-studied area of robust estimators in statisticss=-=[30]-=-. An estimator is said to be robust with a breakdownspoint of a if the estimator's value is immune to the place-sment of any fraction of up to c~ outliers, no matter howsfar they lie from the rest of ... |

1140 | Integer and Combinatorial Optimization - Nemhauser, Wolsey - 1988 |

658 | A threshold of ln n for approximating set cover - Feige - 1998 |

604 |
Approximation Algorithms for NP-Hard Problems
- Hochbaum, editor
- 1997
(Show Context)
Citation Context ...lem for any given value of p.sSince the robust k-center problem is a generaliza-stion of the (non-robust) k-center, the lower bound ofs2 - c (assuming P ¢ NP) for the approximation ratiosmentioned in =-=[16]-=- also holds for the robust case. Thesfollowing theorem proves a hardness of 3 - c on thesperformance of any approximation algorithm for the ro-sbust k-center problem (with forbidden centers) for input... |

358 | A general approximation technique for constrained forest problems
- Goemans, Williamson
- 1995
(Show Context)
Citation Context ...escost to its nearest facility or to pay the penalty.sSetting the penalties to cx~ gives the standard for-smulation. (This notion has been studied earlier insthe context of TSP and Steiner trees, see =-=[11, 10]-=-sand references therein.)s2-center solution (k=2)s®s0s2-center robust solution (k=2, p= 11) ®s@sFigure 1: Example to illustrate how robust measuresslead to better clustering solutions.sThese two modif... |

336 | Polynomial time approximation schemes for Euclidean TSP and other geometric problems
- Arora
- 1998
(Show Context)
Citation Context ...eck-TSP problemsas well. Rather than asking for a rain-max tour visitingsall n locations, we would like to visit only a specifiedsnumber of locations.sFor Euclidean instances of problems uch as k-TSPs=-=[1]-=-, k-medians [2], etc., we note that the PTAS's cansall be extended to obtain PTAS's for robust versions ofsthese problems.s2 LowersBoundssFeder and Greene [8] showed that if k is part of thesinput, th... |

326 | Approximation algorithms for metric facility location and k-median problems using the primal-dual schema and Lagrangian relaxation
- Jain, Vazirani
(Show Context)
Citation Context ...(rather thansthe sum of distances) from any facility to its closest opensfacility. All of these optimization problems are NP-shard, and polynomial time approximation algorithmsshave been studied (see =-=[17, 4, 5, 6, 31, 19, 13, 15, 21, 23]-=-).sA significant shortcoming of these simple formula-stions is that a few very distant clients, called ouLliers,scan exert a disproportionately strong influence over thesfinal solution. This is clearl... |

272 | Approximation algorithms for facility location problems
- Shmoys
- 2000
(Show Context)
Citation Context ...(rather thansthe sum of distances) from any facility to its closest opensfacility. All of these optimization problems are NP-shard, and polynomial time approximation algorithmsshave been studied (see =-=[17, 4, 5, 6, 31, 19, 13, 15, 21, 23]-=-).sA significant shortcoming of these simple formula-stions is that a few very distant clients, called ouLliers,scan exert a disproportionately strong influence over thesfinal solution. This is clearl... |

220 | A constant-factor approximation algorithm for the k-median problem
- Charikar, Guha, et al.
(Show Context)
Citation Context ...(rather thansthe sum of distances) from any facility to its closest opensfacility. All of these optimization problems are NP-shard, and polynomial time approximation algorithmsshave been studied (see =-=[17, 4, 5, 6, 31, 19, 13, 15, 21, 23]-=-).sA significant shortcoming of these simple formula-stions is that a few very distant clients, called ouLliers,scan exert a disproportionately strong influence over thesfinal solution. This is clearl... |

209 | Improved combinatorial algorithms for facility location problems - Charikar, Guha |

190 | Greedy strikes back: Improved facility location algorithms
- Guha, Khuller
- 1999
(Show Context)
Citation Context |

174 |
Optimal algorithms for approximate clustering
- Feder, Greene
- 1988
(Show Context)
Citation Context ...lidean instances of problems uch as k-TSPs[1], k-medians [2], etc., we note that the PTAS's cansall be extended to obtain PTAS's for robust versions ofsthese problems.s2 LowersBoundssFeder and Greene =-=[8]-=- showed that if k is part of thesinput, then the k-center problem (i.e., the non-robustsversion) is NP-hard to approximate to within a factor ofs1.822 even for points in the plane (with the Euclideans... |

150 | Analysis of a local search heuristic for facility location problems - Korupolu, Plaxton, et al. |

132 | The budgeted maximum coverage problem
- Khuller, Moss, et al.
- 1999
(Show Context)
Citation Context ...forbidden centers cannot be approximated toswithin a factor of 3 - e unless P : NP.sProof. Our proof uses a result on the maximum coveragesproblem proved independently by Feige [9] and Khullerset al. =-=[20]-=-. The problem is defined as follows: givensa collection S of sets over a domain B of elements,sfinding a subset SS p C S of cardinality k that maximizessthe number of elements covered from B is a NP-h... |

123 | Improved approximation algorithms for the uncapacitated facility location problem
- Chudak
- 1998
(Show Context)
Citation Context |

117 | Approximation schemes for Euclidean k-medians and related problems
- Arora, Raghavan, et al.
- 1998
(Show Context)
Citation Context ...sas well. Rather than asking for a rain-max tour visitingsall n locations, we would like to visit only a specifiedsnumber of locations.sFor Euclidean instances of problems uch as k-TSPs[1], k-medians =-=[2]-=-, etc., we note that the PTAS's cansall be extended to obtain PTAS's for robust versions ofsthese problems.s2 LowersBoundssFeder and Greene [8] showed that if k is part of thesinput, then the k-center... |

98 |
A heuristic program for locating warehouses
- Kuehn, Hamburger
- 1963
(Show Context)
Citation Context ... with this added constraint.s1 In t roduct ionsThe facility location problem and the related clusteringsproblems, k-median and k-center, are widely studied insoperations research and computer science =-=[3, 7, 22, 24,s32]-=-. Typically in problems of this type, we are givensan n-vertex graph whose edge weights define a distancesmetric. Let eij denote the distance between odes i andsj .sIn the (uncapacitated) facility loc... |

92 |
A unified approach to approximation algorithms for bottleneck problems
- Hochbaum, Shmoys
- 1986
(Show Context)
Citation Context |

81 |
Heuristics for the fixed cost median problem
- Hochbaum
- 1982
(Show Context)
Citation Context |

71 | Approximation algorithms for geometric median problems
- Lin, Vitter
- 1992
(Show Context)
Citation Context |

66 | A Note on the Prize Collecting Traveling Salesman Problem
- Bienstock, Goemans, et al.
- 1993
(Show Context)
Citation Context ...escost to its nearest facility or to pay the penalty.sSetting the penalties to cx~ gives the standard for-smulation. (This notion has been studied earlier insthe context of TSP and Steiner trees, see =-=[11, 10]-=-sand references therein.)s2-center solution (k=2)s®s0s2-center robust solution (k=2, p= 11) ®s@sFigure 1: Example to illustrate how robust measuresslead to better clustering solutions.sThese two modif... |

37 |
A simple heuristic for the p-center problem
- Dyer, Frieze
- 1985
(Show Context)
Citation Context ...on of the population to which we are willing tosdeny service. An example is shown in Figure 1.sNote that by using the notion of "weights" in thesK-center problem one can reduce the effect of outlierss=-=[12, 29]-=- and design constant factor approximation al-sgorithms. (The definition of the distance function issmodified to be a weighted istance function, by multi-splying the weight of the node with its distanc... |

25 |
On finding integer solutions to linear programs
- Balinski
- 1966
(Show Context)
Citation Context ... with this added constraint.s1 In t roduct ionsThe facility location problem and the related clusteringsproblems, k-median and k-center, are widely studied insoperations research and computer science =-=[3, 7, 22, 24,s32]-=-. Typically in problems of this type, we are givensan n-vertex graph whose edge weights define a distancesmetric. Let eij denote the distance between odes i andsj .sIn the (uncapacitated) facility loc... |

25 | A Working Model for Plant Numbers and Locations - Stollsteimer - 1963 |

22 |
A heuristic for the p-center problem in graphs
- Plesnik
- 1987
(Show Context)
Citation Context ...on of the population to which we are willing tosdeny service. An example is shown in Figure 1.sNote that by using the notion of "weights" in thesK-center problem one can reduce the effect of outlierss=-=[12, 29]-=- and design constant factor approximation al-sgorithms. (The definition of the distance function issmodified to be a weighted istance function, by multi-splying the weight of the node with its distanc... |

17 | A practical approximation algorithm for the lms line estimator - Mount, Netanyahu, et al. - 1997 |

12 | Efficient randomized algorithms for the repeated median line estimator - Matouek, Mount, et al. - 1993 |

11 |
Plant location under economies-of- scale-decentralization and computation
- Manne
- 1964
(Show Context)
Citation Context ... with this added constraint.s1 In t roduct ionsThe facility location problem and the related clusteringsproblems, k-median and k-center, are widely studied insoperations research and computer science =-=[3, 7, 22, 24,s32]-=-. Typically in problems of this type, we are givensan n-vertex graph whose edge weights define a distancesmetric. Let eij denote the distance between odes i andsj .sIn the (uncapacitated) facility loc... |

7 | Efficient randomized algorithms for the repeated median line estimator. Algorithmica - Matousek, Mount, et al. - 1998 |

4 | Efficient algorithms for robust point pattern matching and applications to image registration - Mount, Netanyahu, et al. - 1997 |

3 |
A threshold of Inn for approximating set cover
- Feige
- 1998
(Show Context)
Citation Context ...-center prob-slem with forbidden centers cannot be approximated toswithin a factor of 3 - e unless P : NP.sProof. Our proof uses a result on the maximum coveragesproblem proved independently by Feige =-=[9]-=- and Khullerset al. [20]. The problem is defined as follows: givensa collection S of sets over a domain B of elements,sfinding a subset SS p C S of cardinality k that maximizessthe number of elements ... |

2 | The uncapacitated facility location problem", Discrete Location Theory (ed: Mirchandani and Francis - Cornuejols, Nemhauser, et al. - 1990 |

2 |
a silent e
- Loud, clear
(Show Context)
Citation Context ...ay be economicallysessential to ignore very distant outliers. For example,sKmart claims to provide service to 88% of the USspopulation within a radius of 6 miles using the currentsset of Kmart stores =-=[14]-=-. Clearly, if they tried toscover the entire population of the country, either thescovering radius or the number of stores would needsto be significantly higher. The simple formulationssof facility lo... |

2 |
Algorithms for Clusterin 9Data
- Jain, Dubes
- 1988
(Show Context)
Citation Context ...l ts .sThe robust k-scenter problem is of interest in clustering applications insdata analysis, where erroneous data is present or wheresit is desirable to ignore less significant subsets of thesdata =-=[18]-=-.sLowersBounds :sIf the input points are in ~d (with thesEuclidean metric), then it is NP-hard to approxi-smate the optimal cost (for all variants) to within asfactor of 1.822. If the input points in ... |

2 | Integer and Combina- torial Optimization - Nemhauser, Wolsey - 1990 |

1 |
Improved combinatorial - gorithms for the facility location and k-median prob- lems
- Charikar, Guha
- 1999
(Show Context)
Citation Context |

1 |
The uncapacitated facility location problem", Dis- crete Location Theory (ed: Mirchandani and Francis
- Cornuejols, Nemhauser, et al.
- 1990
(Show Context)
Citation Context |

1 |
Anal- ysis of a local search euristic for facility location prob- lems
- Korupolu, Plaxton, et al.
- 1998
(Show Context)
Citation Context |