Linear code properties, not preserved by equivalence

118 Views Asked by At

Let $C\leq F^n$ be a linear code of length $n$, over the field $F$.

We say that the linear code $D$ is equivalent to $C$ if it differs by a fixed reordering of the vector coordinates. I.e., there exists $\sigma \in S_n$, such that $\sigma:C \to D$ is a bijection. Where the action is defined $\sigma (c_1,\dots ,c_n)=(c_{\sigma(1)},\dots c_{\sigma(n)})$.

Now, this is a natural definition of equivalence as it preserves many important properties of codes, for example, minimum distance.

My question is then, are there any interesting properties $X$ or theorems of the following form:

"Given a linear code $C$, there exists an equivalent code $D$ such that $X$ property holds"?

I.e., is there anything nice or useful that may not be true for our original code, but will be for one that is equivalent?

Thanks.

1

There are 1 best solutions below

2
On

Some codes (typically, cyclic codes) have specific burst error detection properties.

For example, consider the cyclic Hamming $(7,4)$ code given by $g(x)=x^3+x+1$. Among its properties:

  1. It has distance $d=3$, hence it can detect $t=2$ errors inside a block, and can correct $1$ error.

  2. It's cyclic. Its codewords have a maximum span length of that of $g(x)$, i.e., $\ell=4$. Hence (in addition of the properties above) it can detect any burst error of length $\ell-1=3$ or less.

Properties $1$ are invariant under equivalence. Properties $2$ are not.

More in general, the probability of decoding error, assuming the model of independent errors, depends only on the weight distribution of the codewords (weight enumerator function) - and this is contant for equivalent codes. But if the errors are not independent the above is not true, and two equivalent codes can have different probability of decoding error.