## Practical Methods for Shape Fitting and Kinetic Data Structures using Core Sets (2004)

### Cached

### Download Links

- [biogeometry.duke.edu]
- [www.cs.duke.edu]
- [www.cs.duke.edu]
- [www.cs.duke.edu]
- [biogeometry.duke.edu]
- [www.cs.duke.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | In Proc. 20th Annu. ACM Sympos. Comput. Geom |

Citations: | 27 - 8 self |

### BibTeX

@INPROCEEDINGS{Yu04practicalmethods,

author = {Hai Yu and Pankaj K. Agarwal and Raghunath Poreddy and Kasturi R. Varadarajan},

title = {Practical Methods for Shape Fitting and Kinetic Data Structures using Core Sets},

booktitle = {In Proc. 20th Annu. ACM Sympos. Comput. Geom},

year = {2004},

pages = {263--272}

}

### Years of Citing Articles

### OpenURL

### Abstract

The notion of ε-kernel was introduced by Agarwal et al. [5] to set up a unified framework for computing various extent measures of a point set P approximately. Roughly speaking, a subset Q ⊆ P is an ε-kernel of P if for every slab W containing Q, the expanded slab (1 + ε)W contains P. They illustrated the significance of ε-kernel by showing that it yields approximation algorithms for a wide range of geometric optimization problems. We present a simpler and more practical algorithm for computing the ε-kernel of a set P of points in R d. We demonstrate the practicality of our algorithm by showing its empirical performance on various inputs. We then describe an incremental algorithm for fitting various shapes and use the ideas of our algorithm for computing ε-kernels to analyze the performance of this algorithm. We illustrate the versatility and practicality of this technique by implementing approximation algorithms for minimum enclosing cylinder, minimum-volume bounding box, and minimum-width annulus. Finally, we show that ε-kernels can be effectively used to expedite the algorithms for maintaining extents of moving points. 1

### Citations

773 | An optimal algorithm for approximate nearest neighbor searching in fixed dimensions
- Arya, Mount, et al.
- 1998
(Show Context)
Citation Context ...he convex hull of Q ′ by solving an abstract optimization problem [19]. In practice our algorithm may take advantage of known approximate nearest-neighbor searching data structures, such as BBD-trees =-=[10]-=-, and the running time would be bounded by O(n log n + log n/ε (3d−1)/2 ), without computing an (ε/2)kernel first. In fact, using a data structure described in [16], the running time can be improved t... |

387 | Applications of random sampling in computational geometry
- Clarkson, Shor
- 1989
(Show Context)
Citation Context ...lgorithm can be somewhat slower than the latter. The idea of our above incremental approach is far from new: it can be found in many randomized incremental algorithms in computational geometry (e.g., =-=[18, 17]-=-), or even earlier in the column generation method for linear programming [21]. In the context of shape fitting, a similar incremental algorithm was proposed by Bǎdoiu et al. [14] for computing the sm... |

231 | Data Structures for Mobile Data
- Basch, Guibas, et al.
- 1997
(Show Context)
Citation Context ... been work on maintaining extent measures of a set of moving points [4]. In the latter, the goal is to maintain the extent measure as the points move, e.g., using the kinetic data structure framework =-=[12]-=-. ∗ A preliminary version of the paper appeared in Proceedings of the 20th Annual ACM Symposium on Computational Geometry, 2004, pp. 263–272. Research by the first two authors is supported by NSF unde... |

192 | ANN: Library for approximate nearest neighbor searching. ftp.cs.umd.edu/pub/faculty/mount/ann
- Mount, Arya
(Show Context)
Citation Context ...then compute ε-kernels for a set of linear functions, polynomials, or their roots. We have implemented our ε-kernel algorithm, using the ANN library for answering approximate nearest-neighbor queries =-=[30]-=-, and tested on a variety of inputs in dimension up to 8. The empirical results show that our algorithm works extremely well in low dimensions (≤ 4) both in terms of the 1 Chan [16] has independently ... |

187 |
On constructing minimum spanning trees in k-dimensional spaces and related problems
- Yao
- 1982
(Show Context)
Citation Context ...εα ≤ 1/2. It is well known that one can construct a set I of O(1/λ d−1 ) = O(1/ε (d−1)/2 ) points on the sphere S such that for any point x on S, there exists a point y ∈ I such that �x − y� ≤ λ (see =-=[32]-=-). We process P into a data structure that can answer ε-approximate nearest-neighbor queries. For a query point q, let ϕ(q) be the point of P returned by this data structure. For each point y ∈ I, we ... |

112 | Approximate clustering via core-sets
- Bădoiu, Har-Peled, et al.
- 2002
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 24, 25, 26, 27, 28]-=-. These approaches compute a subset Q ⊆ P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |

102 | A Las Vegas algorithm for linear and integer programming when the dimension is small
- CLARKSON
- 1995
(Show Context)
Citation Context ...m takes O(1/ε (d−1)/2 ) iterations), our empirical results show that in practice the algorithm converges in very few iterations. Our approach is reminiscent of Clarkson’s linear-programming algorithm =-=[17]-=-. We illustrate the versatility of the technique by implementing approximation algorithms for minimum enclosing cylinder, minimum-volume bounding box, and minimum-width annulus. We compare our increme... |

96 | K.R.: Approximating extent measures of points
- Agarwal, Har-Peled, et al.
- 2004
(Show Context)
Citation Context ...d Kinetic Data Structures using Coresets ∗ Hai Yu † Pankaj K. Agarwal † Raghunath Poreddy ‡ Kasturi R. Varadarajan ‡ December 10, 2006 Abstract The notion of ε-kernel was introduced by Agarwal et al. =-=[5]-=- to set up a unified framework for computing various extent measures of a point set P approximately. Roughly speaking, a subset Q ⊆ P is an ε-kernel of P if for every slab W containing Q, the expanded... |

96 |
A linear programming approach to the cutting-stock problem
- Gilmore, Gomory
- 1963
(Show Context)
Citation Context ... approach is far from new: it can be found in many randomized incremental algorithms in computational geometry (e.g., [18, 17]), or even earlier in the column generation method for linear programming =-=[21]-=-. In the context of shape fitting, a similar incremental algorithm was proposed by Bǎdoiu et al. [14] for computing the smallest enclosing ball of a point set and by Kumar and Yildirim [28] for comput... |

87 | A replacement for Voronoi diagrams of near linear size
- Har-Peled
- 2001
(Show Context)
Citation Context ...hat it might still be useful as a first prune in practice, and the running time may be improved by a better tradeoff between the preprocessing and query time (e.g., using approximate Voronoi diagrams =-=[9, 22]-=- for answering approximate nearest-neighbor queries). Our algorithm still performed well on the input clustered in dimensions 6 and 8. Figure 2 shows that, as the size of the computed ε-kernel becomes... |

77 | Efficiently approximating the minimum-volume bounding box of a point set in three dimensions
- Barequet, Har-Peled
(Show Context)
Citation Context ...nclosing cylindrical shell requires O(n 5 ) time [3]. Consequently, attention has shifted to developing approximation algorithms, and several approximation algorithms for specific problems exist; see =-=[1, 3, 11, 15, 33]-=- and the references therein. Although these algorithms are tailored to specific problems, they rely on similar techniques. A natural open question is thus whether a unified framework exists for comput... |

64 | Faster core-set constructions and data-stream algorithms in fixed dimensions
- Chan
- 2006
(Show Context)
Citation Context ...t-neighbor queries [30], and tested on a variety of inputs in dimension up to 8. The empirical results show that our algorithm works extremely well in low dimensions (≤ 4) both in terms of the 1 Chan =-=[16]-=- has independently observed a similar simplification, and developed an algorithm for computing an ε-kernel of size O(1/ε (d−1)/2 ) whose running time is O(n + 1/ε d−3/2 ). Our focus is on the empirica... |

60 | Geometric approximation via coresets - Agarwal, Har-Peled, et al. - 2005 |

60 |
Approximating the diameter, width, smallest enclosing cylinder, and minimum-width annulus
- Chan
(Show Context)
Citation Context ...nclosing cylindrical shell requires O(n 5 ) time [3]. Consequently, attention has shifted to developing approximation algorithms, and several approximation algorithms for specific problems exist; see =-=[1, 3, 11, 15, 33]-=- and the references therein. Although these algorithms are tailored to specific problems, they rely on similar techniques. A natural open question is thus whether a unified framework exists for comput... |

47 | Coresets for k-means and k-median clustering and their applications
- Har-Peled, Mazumdar
- 2004
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 24, 25, 26, 27, 28]-=-. These approaches compute a subset Q ⊆ P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |

46 | A subexponential algorithm for abstract optimization problems
- Gärtner
- 1995
(Show Context)
Citation Context ...s an (ε/2)-kernel Q ′ of size O(1/εd−1 ), and the second stage computes an (ε/3)-kernel Q of Q ′ of size O(1/ε (d−1)/2 ). The second stage invokes O(1/ε (d−1)/2 ) times Gärtner’s randomized algorithm =-=[19]-=- (or a deterministic counterpart) that computes the face of the convex hull conv(Q ′ ) closest to a given point lying outside conv(Q ′ ). The deterministic counterparts are rather complicated and even... |

42 | On range searching with semialgebraic sets, Discrete Comput - Agarwal, Matouˇsek - 1994 |

42 | Optimal core-sets for balls
- Bădoiu, Clarkson
- 2008
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 24, 25, 26, 27, 28]-=-. These approaches compute a subset Q ⊆ P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |

40 | Smaller core-sets for balls
- Bădoiu, Clarkson
- 2003
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 22, 23, 24, 25, 26]-=-. These approaches compute a subset Q ⊆ P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |

34 | Approximate minimum enclosing balls in high dimensions using core-sets
- Kumar, Mitchell, et al.
- 2003
(Show Context)
Citation Context |

33 | Line transversals of balls and smallest enclosing cylinders in three dimensions
- Agarwal, Aronov, et al.
- 1997
(Show Context)
Citation Context ...rithm for computing the smallest enclosing cylinder. The exact algorithms for computing the smallest enclosing cylinder are not only expensive (the best known exact algorithm requires near-cubic time =-=[2]-=-), but also involve complicated machineries such as parametric search and computing the roots of high-degree polynomials. Therefore we implemented the approximation algorithm by Agarwal et al. [2], wh... |

33 | Approximation algorithms for k-line center
- Agarwal, Procopiuc, et al.
- 2002
(Show Context)
Citation Context |

28 | Y.: Shape Fitting with Outliers
- Har-Peled, Wang
- 2004
(Show Context)
Citation Context |

27 | Space-Efficient Approximate Voronoi Diagrams
- Arya, Malamatos, et al.
- 2002
(Show Context)
Citation Context ...hat it might still be useful as a first prune in practice, and the running time may be improved by a better tradeoff between the preprocessing and query time (e.g., using approximate Voronoi diagrams =-=[9, 22]-=- for answering approximate nearest-neighbor queries). Our algorithm still performed well on the input clustered in dimensions 6 and 8. Figure 2 shows that, as the size of the computed ε-kernel becomes... |

26 | Smaller coresets for k-median and k-means clustering. Discrete Computational Geometry, 37, 3–19. Oper Res
- Har-Peled, Kushal
- 2007
(Show Context)
Citation Context |

25 | Minimum-volume enclosing ellipsoids and core sets
- Kumar, Yıldırım
(Show Context)
Citation Context |

24 | A practical approach for computing the diameter of a point set
- Har-Peled
- 2001
(Show Context)
Citation Context ...rentheses. Running time is measured in seconds. Minimum-volume box. We used an approximation algorithm (bhbx) by Barequet and Har-Peled [11] for computing Aopt, whose implementation was obtained from =-=[23]-=-. We also used their constant-factor approximation algorithm to choose a set of four points as the initial subset R. We compared our incremental algorithm (incr) with bhbx on various inputs. The input... |

23 | Approximation and exact algorithms for minimum – width annuli and shells
- Agarwal, Aronov, et al.
- 1999
(Show Context)
Citation Context ...nclosing cylindrical shell requires O(n 5 ) time [3]. Consequently, attention has shifted to developing approximation algorithms, and several approximation algorithms for specific problems exist; see =-=[1, 3, 11, 15, 33]-=- and the references therein. Although these algorithms are tailored to specific problems, they rely on similar techniques. A natural open question is thus whether a unified framework exists for comput... |

22 |
On Khachiyan’s algorithm for the computation of minimum-volume enclosing ellipsoids. Discrete Applied Mathematics, 155(13):1731–1744, 2007. Robert Hildebrand
- Todd, Yıldırım
(Show Context)
Citation Context ...ber of steps for a fixed parameter ε. (2) The subset R in our incremental algorithm always grows as the number of iterations increases. For the smallest enclosing ellipsoid problem, Todd and Yildirim =-=[31]-=- noted a ”dropping” idea that allows for dropping points from the subset R, which may leads to a smaller subset R in practice. It is interesting to see whether the same idea can be adapted for the pro... |

14 | Computing core-sets and approximate smallest enclosing hyperspheres in high dimensions
- Kumar, Mitchell, et al.
- 2003
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 22, 23, 24, 25, 26]-=-. These approaches compute a subset Q ⊆ P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |

13 | Exact and approximation algorithms for minimum-width cylindrical shells, Discrete Comput
- Agarwal, Aronov, et al.
(Show Context)
Citation Context ...lly expensive, e.g., for d = 3, the best known algorithm for computing the smallest simplex containing P requires O(n 4 ) time [33], and the smallest enclosing cylindrical shell requires O(n 5 ) time =-=[3]-=-. Consequently, attention has shifted to developing approximation algorithms, and several approximation algorithms for specific problems exist; see [1, 3, 11, 15, 33] and the references therein. Altho... |

12 |
Maintaining the extent of a moving point set, Discrete Comput
- Agarwal, Guibas, et al.
- 1994
(Show Context)
Citation Context ...raditionally P is assumed to be stationary (insertion/deletion of points at discrete times have been considered), recently there has been work on maintaining extent measures of a set of moving points =-=[4]-=-. In the latter, the goal is to maintain the extent measure as the points move, e.g., using the kinetic data structure framework [12]. ∗ A preliminary version of the paper appeared in Proceedings of t... |

11 | Algorithms for a minimum volume enclosing simplex in three dimensions
- Zhou, Suri
(Show Context)
Citation Context ...uiowa.edu 1sThe exact algorithms for computing extent measures are generally expensive, e.g., for d = 3, the best known algorithm for computing the smallest simplex containing P requires O(n 4 ) time =-=[33]-=-, and the smallest enclosing cylindrical shell requires O(n 5 ) time [3]. Consequently, attention has shifted to developing approximation algorithms, and several approximation algorithms for specific ... |

7 | Approximating minimum volume enclosing ellipsoids using core sets, Unpublished manuscript
- Kumar, Yildirim
- 2003
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 22, 23, 24, 25, 26]-=-. These approaches compute a subset Q ⊆ P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |

3 |
Geometric Models Archive. http://www.cc.gatech.edu/projects/large models
- Large
(Show Context)
Citation Context ...[30] for answering approximate nearest-neighbor queries, and we set its relative error to 0.01. We used three different types of synthetic inputs in dimensions 2–8 and a few large 3D geometric models =-=[29]-=-: (i) points uniformly distributed on a sphere (sphere); (ii) points uniformly distributed on a cylindrical surface (cylinder); (iii) clustered point sets (clustered), consisting of 20 equal-sized clu... |

2 |
An optimal algorithm for approximatenearest neighbor searching in fixed dimensions
- Arya, Mount, et al.
- 1998
(Show Context)
Citation Context ...the convex hull of Q0 by solving an abstract optimization problem [19]. In practice our algorithm may take advantage of known approximate nearest-neighbor searching data structures, such as BBD-trees =-=[10]-=-, and the running time would be bounded by O(n log n + log n/"(3d-1)/2), without computing an ("/2)- kernel first. In fact, using a data structure described in [16], the running time can be improved t... |

1 |
Smallest enclosing ball — fast and robust
- Gärtner
(Show Context)
Citation Context ...lgorithm extends readily to the case where Aopt is replaced by an approximation algorithm.) This algorithm requires a procedure for computing smallest enclosing disks, and we have used Gärtner’s code =-=[20]-=- for this purpose. We chose the initial subset R as follows: let p1 be an arbitrary point from the input, p2 be the point farthest from p1, and p3 be the point farthest from the line p1p2; we set R = ... |

1 |
Approximate clustering via core-sets, Proc. 34th Annu. ACMSympos. Theory Comput
- Badoiu, Har-Peled, et al.
- 2002
(Show Context)
Citation Context ...ng time were previously known for these problems. Many subsequent papers have used a similar approach for other geometric optimization problems, including clustering and other extent-measure problems =-=[6, 8, 13, 14, 24, 25, 26, 27, 28]-=-. These approaches compute a subset Q ` P of small size and solve the underlying optimization problem for Q. The term coreset is now commonly used to refer to such a subset. Although the algorithm by ... |