Scribd
Upload
7 views6 pages

Lab 06 210021110

The document outlines an experiment focused on the differentiation of the continuous function f(x) = e^x using numerical techniques, specifically forward, backward, and central difference approximations. It includes MATLAB code that calculates and displays errors for each method, demonstrating that the central difference method has an error of O(h^2) while the forward and backward methods have an error of O(h). The results are presented in a table and visualized through a plot, along with a discussion on convergence rates.

Uploaded by

Saadman Abeer
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
7 views6 pages

Lab 06 210021110

The document outlines an experiment focused on the differentiation of the continuous function f(x) = e^x using numerical techniques, specifically forward, backward, and central difference approximations. It includes MATLAB code that calculates and displays errors for each method, demonstrating that the central difference method has an error of O(h^2) while the forward and backward methods have an error of O(h). The results are presented in a table and visualized through a plot, along with a discussion on convergence rates.

Uploaded by

Saadman Abeer
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Experiment Name: Differentiation of Continuous Function by Numerical

Techniques

Theory:

Forward Difference Approximation of the First Derivative:

𝑓(𝑥+△𝑋)−𝑓(𝑥)
𝑓'(𝑥) = lim △𝑥
△𝑥 → 0
Where,
△𝑥 = 𝑥𝑖+1 − 𝑥𝑖

Backward Difference Approximation of the First Derivative:

𝑓(𝑥)− 𝑓(𝑥−△𝑋))
𝑓'(𝑥) = lim △𝑥
△𝑥 → 0

Where,
△𝑥 = 𝑥𝑖 − 𝑥𝑖−1

Central difference approximation of the first derivative:

𝑓(𝑥+△𝑋)−𝑓(𝑥−△𝑥)
𝑓'(𝑥) = lim 2△𝑥
△𝑥 → 0

Task :
Write a matlab code which will show that error while using central difference method for
approximating first derivative of function is O(x2) but O(x) if forward or backward difference
method is used.

Code:

f = @(x) exp(x);
f_exact_deriv = @(x) exp(x);

% Evaluation at poin t x=0.5


x0 = 0.5;

% Array of step sizes (decreasing by factor of 2)


n = 8;
h = zeros(n,1);
for i = 1:n
h(i) = 0.1 * 2^(1-i);
end

forward_approx = zeros(n,1);
backward_approx = zeros(n,1);
central_approx = zeros(n,1);
forward_error = zeros(n,1);
backward_error = zeros(n,1);
central_error = zeros(n,1);

% Calculating approximations and errors


exact = f_exact_deriv(x0);
for i = 1:n
% Forward difference
forward_approx(i) = (f(x0 + h(i)) - f(x0)) / h(i);
forward_error(i) = abs(forward_approx(i) - exact);

% Backward difference
backward_approx(i) = (f(x0) - f(x0 - h(i))) / h(i);
backward_error(i) = abs(backward_approx(i) - exact);

% Central difference
central_approx(i) = (f(x0 + h(i)) - f(x0 - h(i))) / (2*h(i));
central_error(i) = abs(central_approx(i) - exact);
end
% Create table to display results
T = table(h, forward_error, backward_error, central_error, ...
'VariableNames', {'h', 'Forward_Error', 'Backward_Error', 'Central_Error'});

% Display table
disp('Error Analysis for Numerical Differentiation:');
disp(T);

% Calculate and display convergence rates


disp('Convergence rates (log2(error ratio)):');
for i = 2:n
rate_forward = log2(forward_error(i-1)/forward_error(i));
rate_backward = log2(backward_error(i-1)/backward_error(i));
rate_central = log2(central_error(i-1)/central_error(i));

if i == n
fprintf('Final convergence rates:\n');
fprintf('Forward: %.2f (expected ~1 for O(h))\n', rate_forward);
fprintf('Backward: %.2f (expected ~1 for O(h))\n', rate_backward);
fprintf('Central: %.2f (expected ~2 for O(h^2))\n', rate_central);
end
end

% Plot errors
figure;
semilogy(h, forward_error, 'bo-', 'LineWidth', 1.5, 'DisplayName', 'Forward');
hold on;
semilogy(h, backward_error, 'r*-', 'LineWidth', 1.5, 'DisplayName', 'Backward');
semilogy(h, central_error, 'gs-', 'LineWidth', 1.5, 'DisplayName', 'Central');
grid on;
xlabel('Step Size h');
ylabel('Absolute Error');
title('Error vs Step Size for f(x) = e^x at x = 0.5');
legend('Location', 'best');
set(gca, 'XDir', 'reverse'); % Reverse x-axis to show decreasing h
hold off;

Output:
Discussion:

In the experiment , we can see the first order differentiation of the function f(x) = e^x
The difference of step size progression starts at 0.1 and halves each time.
Outputs results in a table showing forward error, central error and backward error .

We calculate convergence rates using log2 of error ratios because h is ½ times in each
progression. So log2(error(i-1)/forward_error(i)) gives us the order

You might also like