mirror of
https://github.com/kuhyx/WUT_Computer_Science.git
synced 2026-11-20 07:14:26 +01:00
108 lines
3.2 KiB
Matlab
108 lines
3.2 KiB
Matlab
% ENUME MICHAŁ SZOPIŃSKI
|
|
% PROJECT A NUMBER 62
|
|
% TASK 4
|
|
% https://github.com/Lachcim/szopinski-enume
|
|
|
|
% define matrix of choice
|
|
A = [1, 1, 7, 5, 2;
|
|
1, 8, 5, 4, 4;
|
|
7, 5, 0, 8, 8;
|
|
5, 4, 8, 0, 8;
|
|
2, 4, 8, 8, 1];
|
|
|
|
% define available algorithms
|
|
algorithms = {
|
|
'without shifts', @eigennoshifts;
|
|
'with shifts', @eigenshifts
|
|
};
|
|
|
|
% solve problem using both algorithms
|
|
for alg = 1:size(algorithms, 1)
|
|
[algname, algfunc] = algorithms{alg, :};
|
|
|
|
% find solution and iteration count using the given algorithm
|
|
[solution, iterations, finalMatrix] = algfunc(A);
|
|
disp(['Solution using QR ', algname, ':']);
|
|
disp(solution);
|
|
disp('Final matrix:');
|
|
disp(finalMatrix);
|
|
disp(['Iteration count: ', num2str(iterations)]);
|
|
end
|
|
|
|
% finds the eigenvalues of a matrix using the QR method without shifts
|
|
function [eigenvalues, iterations, A] = eigennoshifts(A)
|
|
% initialize iteration counter
|
|
iterations = 0;
|
|
|
|
while 1
|
|
% converge to eigenvalue diagonal matrix
|
|
[Q, R] = qrdecomp(A);
|
|
A = R * Q;
|
|
iterations = iterations + 1;
|
|
|
|
% iterate until all non-diagonal elements are below the threshold
|
|
nondiag = A - diag(diag(A));
|
|
maxnonzero = max(max(abs(nondiag)));
|
|
if (maxnonzero <= 1e-6); break; end
|
|
end
|
|
|
|
% convert eigenvalue matrix to vector
|
|
eigenvalues = diag(A)';
|
|
end
|
|
|
|
% finds the eigenvalues of a matrix using QR with shifts
|
|
function [eigenvalues, iterations, finalMatrix] = eigenshifts(A)
|
|
% initialize empty output
|
|
eigenvalues = double.empty(1, 0);
|
|
iterations = 0;
|
|
|
|
% consider increasingly smaller sub-matrices
|
|
matsize = size(A, 1);
|
|
while matsize >= 2
|
|
% find one eigenvalue
|
|
while 1
|
|
% find eigenvalue of the lower right corner
|
|
corner = A((matsize - 1):matsize, (matsize - 1):matsize);
|
|
cornereigen = eigenoftwo(corner);
|
|
|
|
% shift and iterate algorithm
|
|
A = A - eye(matsize) * cornereigen;
|
|
[Q, R] = qrdecomp(A);
|
|
A = R * Q + eye(matsize) * cornereigen;
|
|
iterations = iterations + 1;
|
|
|
|
% move on once the row has been zeroed out
|
|
maxnonzero = max(abs(A(matsize, 1:(matsize - 1))));
|
|
if (maxnonzero <= 1e-6)
|
|
% remember discovered eigenvalue
|
|
eigenvalues(size(eigenvalues, 2) + 1) = A(matsize, matsize);
|
|
break;
|
|
end
|
|
end
|
|
|
|
% deflate matrix
|
|
finalMatrix(1:matsize, 1:matsize) = A;
|
|
matsize = matsize - 1;
|
|
A = A(1:matsize, 1:matsize);
|
|
end
|
|
|
|
% include final eigenvalue
|
|
eigenvalues(size(eigenvalues, 2) + 1) = A(1, 1);
|
|
end
|
|
|
|
% finds the eigenvalue of a 2x2 matrix that is closer to the lower right corner
|
|
function eigen = eigenoftwo(A)
|
|
% solve characteristic equation of matrix to find its eigenvalues
|
|
delta = (A(1) + A(4)) ^ 2 - 4 * (A(1) * A(4) - A(2) * A(3));
|
|
sqrtdelta = sqrt(delta);
|
|
eigen1 = ((A(1) + A(4)) - sqrtdelta) / 2;
|
|
eigen2 = ((A(1) + A(4)) + sqrtdelta) / 2;
|
|
|
|
% return value closer to lower right corner
|
|
if abs(A(4) - eigen1) < abs(A(4) - eigen2)
|
|
eigen = eigen1;
|
|
else
|
|
eigen = eigen2;
|
|
end
|
|
end
|