Author: Leonid P. Pryadko & Weilei Zeng
Date: 2024-08-01
-This file (introduction.md) is used as the main page for the doxygen documentation.
Inline formula: here \( a \) is random variable. \(J \in {-1,1} \), and (display mode)
-\[ x \in (0,\infty) \] -
-Then one can compute \(y\) as (display mode) $$ y=a \exp(-Jx) $$
-$single dollar sign$ is not supported for inline formulas
-A Python function can be defined as
The program implements two algorithms for calculating the distance of a classical or quantum CSS binary code:
For a classical (binary linear) code, only matrix H, the parity-check matrix, should be specified.
For a quantum CSS code, matrix H=Hx and either G=Hz or L=Lx matrices are needed.
All matrices with entries in GF(2) should have the same number of columns, n, and obey the following orthogonality conditions: $$H_XH_Z^T=0,\quad H_XL_Z^T=0,\quad L_XH_Z^T=0,\quad L_XL_Z^T=I,$$ where $I$ is an identity matrix. Notice that the latter identity is not required; it is sufficient that Lx and Lz matrices have the same full row rank =k, the dimension of the code, each row of Lx has a non-zero scalar product with a row of Lz, and vice versa.
Given the error model, i.e., the matrices $H=H_x$, $L=L_x$ ($L$ is empty for a classical code), the program searches for smallest-weight binary codewords $c$ such that $Hc=0$, $Lc\neq0$.
It repeatedly calculates reduced row echelon form of H, with columns taken in random order, which uniquely fixes the information set (non-pivot columns). Generally, column permutation and row reduction gives $H=U\,(I|A)\,P$, where $U$ is invertible, $P$ is a permutation matrix, $I$ is the identity matrix of size given by the rank of $H$, and columns of $A$ form the information set of the corresponding binary code. The corresponding codewords can be drawn as the rows of the matrix $(A^T|I')\,P$. Because of the identity matrix $I'$, the distribution of such vectors is tilted toward smaller weight, which qualitatively explains why it works.
steps the number of RW decoding steps (the number of information sets to be constructed).The program tries to construct a codeword recursively, by starting with a non-zero bit in a position i in the range from $0$ to $n-1$, where $n$ is the number of columns, and then recursively adding the additional bits in the support of unsatisfied checks starting from the top. The complexity to enumerate all codewords of weight up to $w$ can be estimated as $n\,(\Delta-1)^{w-1}$, where $\Delta$ is the maximum row weight.
Additional command-line parameters relevant for this method:
@@ -121,7 +103,7 @@wmax the maximum size of the connected cluster.start the position to start the cluster. In this case only one starting position i=start will be used. This is useful, e.g., if the code is symmetric (as, e.g., for cyclic codes).For help, just run ./dist_m4ri -h or ./dist_m4ri --help. This shows the following
The program is intended for use with recent gcc compilers under linux. Download the distribution from github then run from the dist-m4ri/src directory sh make -j all This should compile the executable dist_m4ri.
The program uses m4ri library for binary linear algebra. To install under Ubuntu, run
This file (introduction.md) is used as the main page for the doxygen documentation.
Inline formula: here \( a \) is random variable. \(J \in {-1,1} \), and (display mode)
+\[ x \in (0,\infty) \] +
+Then one can compute \(y\) as (display mode) $$ y=a \exp(-Jx) $$
+$single dollar sign$ is not supported for inline formulas
+A Python function can be defined as