Scribd
Upload
37 views

Divide and Conquer

The document describes an algorithm for finding the closest pair of points among a set of points in d-dimensional space. It uses a divide and conquer approach, recursively splitting the points into halves based on their x-coordinates, and finding the closest pairs within each half. It then only needs to check pairs that are within a narrow strip between the halves. This results in an overall runtime of O(n log n) for two dimensions and O(n logd-1 n) for higher dimensions. The document also presents an algorithm that uses grid placement and nearest neighbor searching to find the closest pair in O((c√d)d n) time when the closest distance is known to be between t and ct.

Uploaded by

nabilshf
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
37 views

Divide and Conquer

The document describes an algorithm for finding the closest pair of points among a set of points in d-dimensional space. It uses a divide and conquer approach, recursively splitting the points into halves based on their x-coordinates, and finding the closest pairs within each half. It then only needs to check pairs that are within a narrow strip between the halves. This results in an overall runtime of O(n log n) for two dimensions and O(n logd-1 n) for higher dimensions. The document also presents an algorithm that uses grid placement and nearest neighbor searching to find the closest pair in O((c√d)d n) time when the closest distance is known to be between t and ct.

Uploaded by

nabilshf
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

October 30, 2003 Lecture 17: Closest Pair 1

Closest Pair
Piotr Indyk
October 30, 2003 Lecture 17: Closest Pair 2
Closest Pair
Find a closest pair among p
1
p
n
R
d
Easy to do in O(dn
2
) time
For all p
i
p
j
, compute ||p
i
p
j
|| and choose
the minimum
We will aim for better time, as long as d is
small
For now, focus on d=2
October 30, 2003 Lecture 17: Closest Pair 3
Divide and conquer
Divide:
Compute the median of x-
coordinates
Split the points into P
L
and
P
R
, each of size n/2
Conquer: compute the
closest pairs for P
L
and P
R
Combine the results (the
hard part)
October 30, 2003 Lecture 17: Closest Pair 4
Combine
Let k=min(k
1
,k
2
)
Observe:
Need to check only pairs which
cross the dividing line
Only interested in pairs within
distance < k
Suffices to look at points in
the 2k-width strip around the
median line
k
1
k
2
2k
October 30, 2003 Lecture 17: Closest Pair 5
Scanning the strip
Sort all points in the strip
by their y-coordinates,
forming q
1
q
r
, r n.
Let y
i
be the y-coordinate
of q
i
For i=1 to r
j=i-1
While y
i
-y
j
< d
Check the pair q
i
,q
j
j:=j-1
d
October 30, 2003 Lecture 17: Closest Pair 6
Analysis
Correctness: easy
Running time is more
involved
Can we have many
q
j
s that are within
distance k from q
i
?
No
Proof by packing
argument
k
October 30, 2003 Lecture 17: Closest Pair 7
Analysis, ctd.
Theorem: there are at most 7
q
j
s such that y
i
-y
j
k.
Proof:
Each such q
j
must lie either
in the left or in the right k k
square
Within each square, all
points have distance
distance k from others
We can pack at most 4 such
points into one square, so
we have 8 points total (incl.
q
i
)
q
i
October 30, 2003 Lecture 17: Closest Pair 8
Packing bound
Proving 4 is not obvious
Will prove 5
Draw a disk of radius k/2
around each point
Disks are disjoint
The disk-square intersection
has area (k/2)
2
/4 = /16 k
2
The square has area k
2
Can pack at most 16/ 5.1
points
October 30, 2003 Lecture 17: Closest Pair 9
Running time
Divide: O(n)
Combine: O(n log n) because we sort by y
However, we can:
Sort all points by y at the beginning
Divide preserves the y-order of points
Then combine takes only O(n)
We get T(n)=2T(n/2)+O(n), so
T(n)=O(n log n)
October 30, 2003 Lecture 17: Closest Pair 10
Higher dimensions
Divide: split P into P
L
and P
R
using the
hyperplane x=t
Conquer: as before
Combine:
Need to take care of points with x in [t-k,t+k]
This is essentially the same problem, but in d-1
dimensions
We get:
T(n,d)=2T(n/2)+T(n,d-1)
T(n,1)=O
d
(1) n
Solves to: T(n,d)=n log
d-1
n
October 30, 2003 Lecture 17: Closest Pair 11
Closest Pair with Help
Given: P={p
1
p
n
} of points from R
d
, such
that the closest distance is in (t,c t]
Goal: find the closest pair
Will give an O((cd)
d
n) time algorithm
Note: by scaling we can assume t=1
October 30, 2003 Lecture 17: Closest Pair 12
Algorithm
Impose a cubic grid onto
R
d
, where each cell is a
1/d1/d cube
Put each point into a
bucket corresponding to
the cell it belongs to
Diameter of each cell is
1, so at most one point
per cell
For each pP, check all
points in cells
intersecting a ball B(p,c)
At most (2dc)
d
such
cells
October 30, 2003 Lecture 17: Closest Pair 13
How to find good t ?
Repeat:
Choose a random point p in P
Let t=t(p)=D(p,P-{p})
Impose a grid with side t< t/(2d), i.e., such
that any pair of adjacent cells has diameter <t
Put the points into the grid cells
Remove all points whose all adjacent cells are
empty
Until P is empty
October 30, 2003 Lecture 17: Closest Pair 14
Correctness
Consider t computed in the last iteration
There is a pair of points with distance t
There is no pair of points with distance t or
less*
We get c=t/t~ 2d
*And never was, if the grids are nested
October 30, 2003 Lecture 17: Closest Pair 15
Running time
Consider t(p
1
)t(p
m
)
An iteration is lucky if t(p
i
) t for at last half
of points p
i
The probability of being lucky is 1/2
Expected #iterations till a lucky one is 2
After we are lucky, the number of points is
m/2
Total expected time = 3
d
times
O(n+n/2+n/4++1)

You might also like