Upper Bounds for Ring Linear Codes

8/6/2019 Upper Bounds for Ring Linear Codes

http://slidepdf.com/reader/full/upper-bounds-for-ring-linear-codes 1/32

Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus

Greferath,Axel Kohnert,

Vitaly

Skachek

Upper Bounds for Ring-Linear Codes

Eimear Byrne1, Marcus Greferath1, Axel Kohnert2,

Vitaly Skachek1

1Claude Shannon Institute andSchool of Mathematical Sciences

University College DublinIreland

2Dept MathematicsUniversity of Bayreuth

Germany

May 19 2009

http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


Vitaly

Skachek

Outline

Codes over finite fields

Code optimalityBounds for codes for the Hamming weight

Ring-linear coding

The homogeneous weight

Bounds on the size of a code for the homogeneous weight

http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


Vitaly

Skachek

Notation

F = GF (q ), q = p m some prime p

R is a finite ring with identity

R̂ := HomZ(R ,C×) the characters on (R , +)

χ ∈ R̂ is a character on (R , +)

C is a code of length n and minimum distance d

http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


Vitaly

Skachek

Using Codes for Error Correction

One parameter that indicates the error-correcting capability of a code is its minimum distance.

http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


VitalySkachek


One parameter that indicates the error-correcting capability of a code is its minimum distance.The higher the minimum distance, the more errors that can bedetected and corrected by the receiver.

http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


VitalySkachek



http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


VitalySkachek



http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


VitalySkachek

Code Optimality

The Main Coding Problem:

1 For fixed length n and minimum distance d , what is themaximum size of any code over R ?i.e., what is AR (n, d )?

2 For a fixed length n and minimum distance d , what is the

maximum size of any linear code over R ?i.e., what is B R (n, d )?

http://find/

http://goback/



Upper Bounds

forRing-Linear

Codes

Eimear Byrne,Marcus


VitalySkachek

Some Distance Functions

Definition (Hamming Metric)

Let u, v ∈ R n. The Hamming distance between u and v is thenumber of components where u and v differ, i.e.

dHam(u, v) = |{i : u i = v i }|

u = [0, 0, 1, 1, 3, 3], v = [1, 2, 2, 1, 1, 3] ∈ Z4

d Ham(u, v) = 4.

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Some Distance Functions

Definition (Lee Metric)

Let u , v ∈ Zm. The Lee distance between u and v is theabsolute value modulo m of u − v , i.e.

dLee(u , v ) = |u − v |m =

u − v if u − v ∈ {0, ..., m/2}v − u otherwise

If u, v ∈ Znm then dLee(u, v) =

i =1..n |u i − v i |m.

u = [0, 0, 1, 1, 3, 3], v = [1, 2, 2, 1, 1, 3] ∈ Z4

d Lee(u, v) = 1 + 2 + 1 + 2 = 6

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Some Bounds for Codes over Finite Fields

Singleton: |C | ≤ Aq (n, d ) ≤ q n−d +1

Hamming: |C | ≤ Aq (n, d ) ≤ q n

V q (n, d −12

),

Plotkin: |C | ≤ Aq (n, d ) ≤ d d −γ n

, γ = q −1q

, if n < d γ

Gilbert-Varshamov: Aq (n, d ) ≥ q n

V q (n,d −1)

Elias-Bassalygo bound

Mc-Eliece-Rodemich-Rumsey-Welch boundLinear Programming bound

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Asymptotic Representations

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Codes over Finite Rings

Definition

An code of length n over R is a nonempty subset of R n. A(left) linear code of length n over R is a left R -submodule of R n.

We will usually assume that R is a finite Frobenius ring.

Many of the foundational results of classical coding theory (e.g.the MacWilliams’ theorems) can be extended to the finite ringcase when R is Frobenius.

[Wood, Honold, Nechaev, Greferath, Schmidt..]

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Finite Frobenius Rings

For a finite ring R , R̂ is an R − R bimodule via

χr (x ) = χ(rx ), r χ(x ) = χ(xr )

for all x , r ∈ R , χ ∈ R̂ .R is a finite Frobenius ring iff

R R R R̂

Then R R̂ = R χ for some (left) generating character χ

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Finite Frobenius Rings

Let R and S be finite Frobenius rings, let G be a finite group.The following are examples of Frobenius rings.

integer residue rings Zm

Galois rings

principal ideal rings

R × S

the matrix ring M n(R )the group ring R [G ]

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Homogeneous Weights

Definition

A weight w : R −→ Q is (left) homogeneous , if w (0) = 0 and

1 If Rx = Ry then w (x ) = w (y ) for all x , y ∈ R .

2 There exists a real number γ such that

y ∈Rx

w (y ) = γ |Rx | for all x ∈ R \ {0}.

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Examples of Homogeneous Weights

ExampleOn every finite field Fq the Hamming weight is a homogeneousweight of average value γ = q −1

q .

Example

On Z4 the Lee weight is homogeneous with γ = 1.

x 0 1 2 3

w Lee(x ) 0 1 2 1

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek


ExampleOn Z10 the following weight is homogeneous with γ = 1:

x 0 1 2 3 4 5 6 7 8 9

whom(x ) 0 1 54 1 5

4 2 54 1 5

4 1

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek


Example

On the ring R of 2 × 2 matrices over GF(2) the weight

w : R −→ R, X →

0 : X = 0,2 : X singular, X = 0,1 : otherwise,

is a homogeneous weight of average value γ = 32 .

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek


Example

On a local Frobenius ring R with q -element residue field theweight

w : R −→ R, x →

0 : x = 0,q

q −1 : x ∈ soc (R ), x = 0,

1 : otherwise,

is a homogeneous weight of average value γ = 1.

Which finite rings admit a homogeneous weight?

Up to the choice of γ , every finite ring admits a uniquehomogeneous weight .

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Homogeneous Weights of FFRs

Theorem (Honold)

Let R be a finite Frobenius ring with generating character χ.

Then the homogeneous weights on R are precisely the functions

w : R −→ R, x → γ

1 −1

|R ×|

u ∈R ×

χ(xu )

where γ is a real number.

http://find/

http://goback/



Upper Bounds

forRing-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Bounds on AR (n, d ) for the Homogeneous Weight

The following bounds have been found for codes over FFRs forthe homogeneous weight.

Sphere-packing (Hamming)

Sphere-covering (Gilbert-Varshamov)

Plotkin-like bounds

Elias-like bounds

Singleton-like boundLinear programming bound

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

A Key Lemma

Lemma

Let C ≤ R R n

be a linear code, and let x ∈ R n

. Then1

|C |

c ∈C

w (x + c ) = γ |supp(C )| +

i ∈supp(C )

w (x i ).

C

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Residual Codes

Definition

Let C ≤R R n, c ∈ R n. Res (C , c ) := {(x i ) : x ∈ C , c i = 0}.Example

Let C be the Z4-linear code generated by

1 0 0 0 3 1 2 10 1 0 0 1 2 3 10 0 1 0 3 3 3 20 0 0 1 2 3 1 1

.

Let c = [0, 0, 0, 2, 0, 2, 2, 2]. Then Res (C , c ) is generated by

1 0 0 30 1 0 10 0 1 3

0 0 0 2

.

R id l C d

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Residual Codes

Theorem

Let C ≤ R R n have minimum homogeneous weight d, and let c ∈ C satisfy (c ) := w Ham < d

γ . Then Res (C , c ) has

length n − (c ),

minimum homogeneous weight d ≥ d − γ(c ),

|Res (C , c )| =|C |

|Rc |and

|C | ≤ |Rc | d − γ(c )d − γ n

.

B d B ( d) f h H W i h

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Bounds on B R (n, d ) for the Homogeneous Weight

Corollary (BGKS)

Let C ≤ R R

n

be a linear code of minimum homogeneous weight d and minimum Hamming weight where ≤ n ≤ d γ

. Then

|C | ≤ |R |d − γ

d − γ n.

B d B ( d) f h H W i h

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek


Corollary (BGKS)

Let C ≤ R R n be a linear code of minimum homogeneous

weight d and minimum Hamming weight where < n ≤ d γ

.Let Q be the maximum size of any minimal ideal of R. Then

|C | ≤ Q d − γ

d − γ n.

A Pl tki O ti l C d

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

A Plotkin Optimal Code

ExampleLet R = F

2×22 . Let C be the length 16m − 1 Simplex Code over

R . Then |C | = 16m,

d = |R |mγ = 16mγ,

:= dHam(C ) = 16m −16m

4=

3

416m.

R has 3 minimal ideals, each of size Q = 4 and so

|C | ≤ Q d − γd − γ n

= 416mγ − 3

4 16mγ

16mγ − (16m − 1)γ = 4

16m

4= 16m.

B d B ( d) f th H W i ht

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek


Singleton-like bounds:

Theorem (BGKS)

Let C ≤ R R n be an [n, d ] linear code and suppose that n ≤ d γ

.Then

n − |R | − 1

|R |

d

γ

≥

log|R | |C | − 1

.

Theorem (BGKS)

Let C be an [n, d ] code over R satisfying n ≤d γ and (C ) < n.

Let P := max{|Ra| : a ∈ R n, Ra ≤ C , (a) < n}. Then

n −

P − 1

P

d

γ

≥ logP |C | − logP |R | .

A Class of MDS Codes

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

A Class of MDS Codes

Example

Let R be a chain ring of length 2. Then R × = R \rad R and|R | = q 2. Let U := R 2\rad R 2, let P := {xR : x ∈ U }. Then

|P| = q 2

+ q .Let C < R R n be the length n := q 2 + q code with 2 × ngenerator matrix whose columns are the distinct elements of P .

Clearly (c ) < n for each c ∈ C .

C is free of rank 2 and the maximal cyclic submodules of C have size P := |R | = q 2.

Let r = logP |C | − 1 = logq 2 q 4 − 1 = 1.

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

Example (cont.)

Setting γ = 1, each word xG of C has weight

w (xG ) =

q 2 + q if x ∈ U q 3

q −1 if x ∈ radR 2,

=⇒ n −

P − 1

P d

= n −

q 2 − 1

q 2(q 2 + q )

= n − q 2 + q − 1 −1

q

= q 2 + q − q 2 − q + 1 = 1 = r .

References

http://find/

http://goback/



Upper Boundsfor

Ring-LinearCodes

Eimear Byrne,Marcus


VitalySkachek

References

E. Byrne, M. Greferath and M. E. O’Sullivan, The linear programming bound for codes over finite Frobenius rings,Designs, Codes and Cryptography, Vol. 42 , 3 (2007), pp.289 - 301.

I. Constantinescu and W. Heise, A metric for codes over

residue class rings of integers , Problemy PeredachiInformatsii 33 (1997), no. 3, 22–28.

M. Greferath and M. E. O’Sullivan, On Bounds for Codes over Frobenius Rings under Homogeneous Weights ,Discrete Mathematics 289 (2004), 11–24.

T. Honold, A characterization of finite Frobenius rings,Arch. Math. (Basel), 76 (2001), 406–415.

J. A. Wood, Duality for modules over finite rings and applications to coding theory, Amer. J. Math. 121

(1999), 555–575.

http://find/

http://goback/

Documents

Upper Bounds for Ring Linear Codes