Let A=[V1, V2 ,..., Vn ] be a list of binary vectors such that the length of Vi for i=1...n is the positive integer number m. For instance, in the following example we have n=6 and m=5.
A := [[1, 1, 1, 0, 0], [0, 1, 0, 1, 1], [1, 0, 1, 1, 1], [0, 1, 1, 1, 0], [1, 1, 0, 1, 0], [0, 1, 1, 1, 1]]
Suppose that ej with j=1...m are vectors of size m such that all entries of ej 's are zero except the jth entry where is equal to 1. Now set S=[e1,e2,...,em]. For example by m=5 we get
S:=[[1, 0, 0, 0, 0], [0, 1, 0, 0, 0], [0, 0, 1, 0, 0], [0, 0, 0, 1, 0], [0, 0, 0, 0, 1]]
First, we choose i,j in [1...m] such that i<>j and then we update S as follows S=[e1,e2,...,em,ei+ej mod 2]. For example, it follows from i=1 and j=2 that
S:=[[1, 0, 0, 0, 0], [0, 1, 0, 0, 0], [0, 0, 1, 0, 0], [0, 0, 0, 1, 0], [0, 0, 0, 0, 1], [1, 1, 0, 0, 0]]
Now consider the kth entry of A which is called Ak. Then we check that what is the minimum number of entries of S so that the summation of these entries mod 2 is equal to Ak. For example, for S=[e1,e2,...,em,e1+e2 mod 2] we get [2, 3, 4, 3, 2, 4] which means the minimum number of entries of the S that are required to be added in order to obtain A1 is 2 and so on. Therefore, for the given example we get
E:=[[1, 2, [2, 3, 4, 3, 2, 4], 18, 7.615773106],
[1, 3, [2, 3, 3, 3, 3, 4], 18, 7.483314774],
[1, 4, [3, 3, 3, 3, 2, 4], 18, 7.483314774],
[1, 5, [3, 3, 3, 3, 3, 4], 19, 7.810249676],
[2, 3, [2, 3, 4, 2, 3, 3], 17, 7.141428429],
[2, 4, [3, 2, 4, 2, 2, 3], 16, 6.782329983],
[2, 5, [3, 2, 4, 3, 3, 3], 18, 7.483314774],
[3, 4, [3, 3, 3, 2, 3, 3], 17, 7.],
[3, 5, [3, 3, 3, 3, 3, 3], 18, 7.348469229],
[4, 5, [3, 2, 3, 3, 3, 3], 17, 7.]];
For instance, the interpretation of [3, 5, [3, 3, 3, 3, 3, 3], 18, 7.348469229] is that if S be [e1,e2,e3,e4,e5,e3+e5 mod 2] then we get [3, 3, 3, 3, 3, 3] where 18 is the summation of [3, 3, 3, 3, 3, 3] and also the number 7.348469229 is obtained from the following command:
MatrixNorm(convert([3, 3, 3, 3, 3, 3], Matrix), Frobenius);
Now we choose Ek for k in [1..nops(E)] such that Ek[4] be minimum over all Ek[4]'s. For example we choose [2, 4, [3, 2, 4, 2, 2, 3], 16, 6.782329983] since 16 is minimum between Ek[4]'s.
There are two points: First one is that if we have Ei and Ej such that Ei[4]=Ej[4] then we choose Ei if Ei[5]>Ej[5] . The second point is that if Ei[4]=Ej[4] and also Ei[5]=Ej[5] then we choose one of them such as the first one. Finally we update the set S from the first two entries of Ek that we have obtained. For instance, the updated S in our example is:
S=[e[1],e[2],e[3],e[4],e[5],S[E[6][1]]+S[E[6][2]] mod 2]=[e[1],e[2],e[3],e[4],e[5],e[2]+e[4] mod 2]
Now we repeat this procedure for the updated S until that in one of the entries of E such as Ek we get Ek[3]=[1,1,..,1].
I have written a procedure in Maple for the mentioned question. But my procedure takes long time to compute when I run it over a list such as A with parameters n=m=64.
I want to kindly request you please modify the following code or suggest another fast procedure for this question.
restart;
with(LinearAlgebra):
with(ListTools):
with(combinat):
BP := proc (A::list)
local n, m, r, S, U, tt, P, E, t, Q, Z, R, k, T, j, PP, i, QQ;
n := nops(A);
m := nops(A[1]);
r := [seq(0, i = 1 .. m)];
r[1] := 1;
S := [seq(Rotate(r, m-i+1), i = 1 .. m)];
unassign('r');
U := [];
tt := 1;
while 0 < tt do
P := choose(nops(S), 2);
E := [];
for t to nops(P) do
Q := P[t];
Z := [];
R := [S[], `mod`(S[Q[1]]+S[Q[2]], 2)];
for k to n do
T := A[k];
for j to nops(R) do
PP := choose(nops(R), j);
for i to nops(PP) do
QQ := PP[i]; r := `mod`(add(R[QQ[i]], i = 1 .. nops(QQ)), 2);
if Occurrences(0, r-T) = m then
Z := [op(Z), j];
i := nops(PP)+1;
j := nops(R)+1
end if;
end do;
unassign('i, QQ, r, PP'):
end do;
end do;
E := [op(E), [Q[], Z, add(Z[i], i = 1 .. nops(Z)), evalf(MatrixNorm(convert(Z, Matrix), Frobenius))]];
unassign('k, Z, Q, R'):
end do;
r := FindMinimalElement([seq(E[i][4], i = 1 .. nops(E))]);
R := [];
for i to nops(E) do
if E[i][4] = r then R := [op(R), E[i]] end if
end do;
T := [FindMaximalElement([seq(R[i][5], i = 1 .. nops(R))], position)];
S := [S[], `mod`(S[R[T[2]][1]]+S[R[T[2]][2]], 2)];
U := [op(U), [R[T[2]][1], R[T[2]][2]]];
if Occurrences(1, R[T[2]][3]) = n then tt := 0 end if;
unassign('r, i, R, T, E')
end do;
return U;
end proc:
A := [[1, 1, 1, 0, 0], [0, 1, 0, 1, 1], [1, 0, 1, 1, 1], [0, 1, 1, 1, 0], [1, 1, 0, 1, 0], [0, 1, 1, 1, 1]];
BP(A);
[[2, 4], [3, 6], [5, 7], [1, 2], [1, 6], [3, 8], [3, 9], [8, 9]]
For more information please see Section 2.1 of the following paper: https://eprint.iacr.org/2019/856.pdf
Thanks in advance for your consideration of this request.