Sei sulla pagina 1di 29

On the Minimum Common

Integer Partition Problem

Author: Xin Chen, Lan Liu,


Zheng Liu, Tao Jiang
Presenter: Lan Liu
Outline
 Introduction
 Problem definitions
 Biological applications
 Approximation of 2-MCIP
 Approximation of k-MCIP
 Conclusion and future work
Problem Definitions
 P(n): given an integer n, a partition is a set of
integers, say {n1,n2,…, nr}, s.t.i=1r ni=n.
Example: given n=4, {2,2} is a P(4);
given n=3, {3} is a P(3).

 IP(S): given a multiset S= {x1, , xm}, an integer


partition is a disjoint union
Example: given S= {3, 3, 4}, {2,2,3,3} is an
IP({3,3,4}).
Observation: S = IP(S)
Examples
 CIP(S1, S2, …, Sk): given multisets S1, S2, …, Sk ,
a common integer partition of all multisets.
Example: given S= {3, 3, 4}, T={2,2,6},
{2,2,3,3} is a CIP(S,T); {1,1,2,2,4} is also a CIP(S,T).
Observation: (1) 9 CIP(S1,…, Sk) $  S1=…=  Sk
(2) |CIP(S1,…, Sk)| ¸ |Si| , i2 {1,..,k}

 MCIP(S1, S2, , Sk): a common integer partition


with the minimum cardinality.
Example: {2,2,3,3} is a MCIP(S,T).
# P(100)=190569292
MCIP is NP-hard
 Subset sum ·P MCIP
 Subset sum problem
Given a set of integer x1, x2,…, xn, s.t. X=ixi, ask if
there is a subset with the sum X/2.
 Reduction to MCIP problem
- Let S={X/2, X/2}, T={x1, x2,…, xn}, find MCIP(S,T).
- If {x1, x2,…, xn} is a MCIP(S,T), the answer is “yes” to
Subset sum problem; otherwise, the answer is “no”.
Biological Applications(1)
 The distance between  Genetic distance between
two strings two genomes
abcdefghijkh
hijkhefgabcd

Minimum Common
Substring Partition
abcdefghijkh
hijkhefgabcd
Biological Applications(2)
 MCIP is a special case of Minimum Common
Substring Partition(MCSP)
MCSP(S,T)
S= aa...a |- aa...a |- aa...a MCIP(S',T')
x1 x2 xn S'= {x1, x2, , xm}
T'= {y1, y2, , yn}
T= aa...a -| aa...a -| aa...a
y1 y2 ym
Outline

 Introduction
 Approximation of 2-MCIP
 Positive results
 Negative results
 Approximation of k-MCIP
 Conclusion and future work
Some basic facts
 |MCIP(S1,S2,…,Sk)|¸
max(|S1|,|S2|,…,|Sk|)

 |MCIP(S,T)|· m+n-1. |
S|=m,|T|=n
Algorithm Analysis
An example: S= {3, 3, 4},T={2,2,6}
Round S T CIP
0 {3,3,4} {2,2,6} ;
1 {1,3,4} {2,6} {2}
2 {3,4} {1,6} {2,1}
3 {2,4} {6} {2,1,1}
4 {4} {4} {2,1,1,2}
5 ; ; {2,1,1,2,4}

 |MCIP(S,T)|· m+n-1
 |MCIP(S,T)| ¸ max(m,n)
 Approximation ratio is 2
Definitions for MRSP(1)
 Related multisets: if S=T and S,T ; , S and T
are a pair of related multisets.
Example:{3, 3, 4,5, 10}
{2, 2, 6,8,7}

 Basic related multisets: if there are no S' ½ S and T'


½ T, s.t. S' and T' are related.
Example: {3, 3, 4,5, 10}
{2, 2, 6,8,7}
Definitions for MRSP(2)
 Maximum Related Multiset Partition problem(MRSP)
Given S and T, partition them into related submultisets
with the maximum cardinality.
Observation: If S, T are (1) {3, 3, 4,5, 10}
a pair of basic related {2, 2, 6,8,7}
multisets, |MRSP|=1. (2) {3, 3, 4,5, 10}
{2, 2, 6,8,7}
(3) {3, 3, 4,5, 10}
{2, 2, 6,8,7}
MRSP $ 2-MCIP
 CIP! RSP
S: {4, 3, 3,5, 10} S: {4, 3, 3,5, 10} S: {4, 3, 3,5, 10}

CIP: {2, 2, 3, 3, 5,3, 7}

T: {2, 2, 6,8,7} T: {2, 2, 6,8,7} T: {2, 2, 6,8,7}

• For each component, •|CIP| ¸ m+n-|RSP|


#edges ¸ #vertices
–1
• |MCIP| ¸ m+n-|RSP|
• Each component is
¸ m+n-|MRSP|
related.
MRSP $2-MCIP
 CIP Ã RSP
 For each related submultisets (S', T'), we run
Greedy_CIP(S', T'), |CIP (S', T')| · |S'|+ |T'|-1
 |CIP| · m+ n- |RSP|
 |MCIP| · |CIP| · m+ n-| MRSP|
MRSP $ 2-MCIP
 |MRSP| = m+n –|MCIP|
 If S, T are a pair of basic related multisets, |MCIP|
= m+n-1, because |MRSP|=1.
 When m+n ¸ 5, |MCIP| =m+n-1¸ 4/5(m+n).

 A new way to solve MCIP


Step1. find MRSP;
Step2. for each basic related submultiset, run
Greedy_CIP(S', T').
Approximate 2-MCIP
 Algorithm intuition:
Step 1. find related submulitsets
mimic MRSP
Step 2. set packing
Step 3. Greedy-CIP
Set Packing Problem(1)
 Set Packing
Given a set of subsets S, find the largest number of
mutually disjoint subsets from S?
Set Packing Problem(2)
 Bad news
- It is NP-hard to find related submultisets of
large size.
- Set packing itself is NP-hard.
 Good news
We can find the small related submultisets and
approximate set packing efficiently.
Approximate 2-MCIP
 Main idea: use different strategies for the
submultisets with different sizes.
Submultiset size Strategies
2 Remove_common_integer
3,4 Approximate_set_packing
5 or more Greedy_CIP

 The approximation ratio is 5/4.


If there are no basic related submultisets with size
smaller than 5, 4/5 (m+n) · |MCIP| · m+n-1.
Outline

 Introduction
 Approximation of 2-MCIP
– Positive results
– Negative results
 Approximation of k-MCIP
 Conclusion and future work
General framework
IP1 IP2
If P1 cannot be approximated
x f(x) within some constant ratio c,
P2 cannot be approximated by
some constant ratio c'.

g(x,y) y

SOLP1(x) SOLP2(f(x))

 Linear Reduction · L
 OPTP2(f(x)) ·  OPTP1(x)
 | OPTP1(x)- g(x,y)| ·|OPTP2(f(x))-y|
Maximum 3DM-3
 Problem Definition
Given a set Dµ X£ Y£ Z, where X, Y and Z are disjoint sets,
and each element occurs in at most three triples, find a
matching with the maximum cardinality.
 Known fact
Maximum 3DM-3 cannot be approximated within some
constant ratio. [Kann91]
X:
X:

Y:
Y:

Z:
Z:
L-reduction(1)
 f: S={4i| i2 X[Y[Z }
T={4i1+4i2+4i3 | (i1,i2,i3)2D}

S: X Y Z

T: D

•OPTMCIP · 70*OPT3DM
L-reduction(2)
 g: - CIP ! RSP
|OPTRSP –SOLRSP| · |OPTMCIP – SOLMCIP |
- RSP ! 3DM
S1: 0 1 ...
1 ...
1 1
•OPT3DM ¼OPTRSP
X Y Z
T1: *
i1
*
i2
* •Each related submultiset
i3
di includes at least one triple
D
|OPT3DM –SOL3DM| · |OPTRSP – SOLRSP |
L-reduction(3)
 There is a constant c s.t. Maximum 3DM-3 cannot
be approximated within c.
 There is a L-reduction s.t.
 OPTMCIP · 70*OPT3DM
 |OPT3DM –SOL3DM| · |OPTMCIP – SOLMCIP|

 There is a constant c' s.t. 2-MCIP cannot be


approximated within c'.
c'<5/4
Outline

 Introduction
 Approximation of 2-MCIP
 Approximation of k-MCIP
 Conclusion and future work
Approximate k-MCIP
 Run Greedy_CIP(S,T) sequentially on S1,S2, …,
Sk.
 |MCIP(S1,S2,…,Sk)| · |S1|+|S2|+…+|Sk|
 |MCIP(S1,S2,…,Sk)| ¸ max(|S1|,|S2|,…,|Sk|)
 Approximation ratio is k
 We can get a {3k(k-1)}/(3k-2)- approximation
by removing the common elements.
Outline

 Introduction
 Approximation of 2-MCIP
 Approximation of k-MCIP
 Conclusion and future work
Upper bound Lower bound
2-MCIP 5/4 APX-hard
k-MCIP (k>2) {3k(k-1)}/(3k-2) APX-hard
Thanks for your time and
attention!

Potrebbero piacerti anche