WUT_Computer_Science/ENUME/references/szopinski-enume/projA/task2gausseli.m
2021-11-10 13:12:25 +01:00

126 lines
4.2 KiB
Matlab

% ENUME MICHAŁ SZOPIŃSKI
% PROJECT A NUMBER 62
% TASK 2
% https://github.com/Lachcim/szopinski-enume
% define maximum feasible equation count (10 * 2^n)
maxeqcount = 5;
% perform calculation for both sub-tasks
for task = 'ab'
% collect error values for different equation counts
errors = zeros(maxeqcount + 1, 2);
% perform calculation for increasing number of equations
for equationspow = 0:maxeqcount
equationcount = 10 * 2 ^ equationspow;
% generate coefficient matrix A and vector b
A = genmatrix(task, equationcount);
b = genvector(task, equationcount);
eqsys = [A, b];
% perform gaussian elimination and back-substitution
[eqsys, lower, upper, permut] = gausseli(eqsys);
result = backsubst(eqsys);
% note the error
errorvector = A * result - b;
error = norm(errorvector);
errors(equationspow + 1, :) = [equationcount, error];
% additional calculations for 10 equations
if equationspow == 0
% print solution
disp(['Solution to subtask ', task ', 10 equations:']);
disp([(1:10)', result, errorvector]);
disp(['Error: ', num2str(error)]);
% residual correction: calculate deltax and correct result
for i = 1:10
% solve system using LU decomposition
intermediate = backsubstflip([lower, permut * errorvector]);
deltax = backsubst([upper, intermediate]);
% apply correction to result
result = result - deltax;
errorvector = A * result - b;
end
% print corrected result
disp('With residual correction:');
disp([(1:10)', result, errorvector]);
disp(['Error: ', num2str(norm(errorvector))]);
end
end
% plot error data
figure;
plot(errors(:, 1), errors(:, 2), '-o');
title(['Linear equation system solution error (subtask ', task, ')']);
xlabel('Equation count');
ylabel('Error');
grid on;
set(gcf, 'PaperPosition', [0 0 6 4]);
set(gcf, 'PaperSize', [6 4]);
print(['report/', task, 'error'], '-dpdf')
end
% performs Gaussian elimination and LU decomposition
function [eqsys, lower, upper, permut] = gausseli(eqsys)
% initialize lower triangular and permutation matrix
lower = zeros(size(eqsys, 1));
permut = eye(size(eqsys, 1));
% for every column, eliminate (size - column) coefficients
for col = 1:size(eqsys, 1)
% partial pivoting - find best row
bestrow = 0;
bestscore = 0;
for row = col:size(eqsys, 1)
score = abs(eqsys(row, col));
if score > bestscore
bestrow = row;
bestscore = score;
end
end
% swap first row and best row
eqsys([col, bestrow], :) = eqsys([bestrow, col], :);
permut([col, bestrow], :) = permut([bestrow, col], :);
for row = (col + 1):size(eqsys, 1)
% calculate reductor constant and perform reduction
reductor = eqsys(row, col) / eqsys(col, col);
eqsys(row, :) = eqsys(row, :) - eqsys(col, :) * reductor;
% add reductor to lower matrix
lower(row, col) = reductor;
% simulate perfect reduction
eqsys(row, col) = 0;
end
% permute the lower matrix
lower([col, bestrow], :) = lower([bestrow, col], :);
lowertmp = lower(col, col);
lower(col, col) = lower(bestrow, col);
lower(bestrow, col) = lowertmp;
end
% return lower and upper triangular matrices
lower = lower + eye(size(lower));
upper = eqsys(:, 1:size(eqsys, 2) - 1);
end
% performs back-substitution for lower triangular matrices
function result = backsubstflip(eqsys)
% flip the equation system for lower triangular matrices
eqcount = size(eqsys, 1);
eqsys(:, 1:eqcount) = rot90(eqsys(:, 1:eqcount), 2);
eqsys(:, eqcount + 1) = flipud(eqsys(:, eqcount + 1));
% perform back-substitution and flip the result
result = flipud(backsubst(eqsys));
end