MATLAB code for Question 2
3 graphs are drawn. First graph is drawn for N from 1 to 10 Second graph is drawn for N from 10 to 50
Third graph is drawn for N from 50 to 150
N = [Link];
theta_1 = 2./N;
theta_2 = 1./N + 100./(N.^2);
figure(1);
plot(N, theta_1, N, theta_2, '.-'), legend('var_est_1', 'var_est_2');
N=[Link];
theta_1 = 2./N;
theta_2 = 1./N + 100./(N.^2);
figure(2);
plot(N, theta_1, N, theta_2, '.-'), legend('var_est_1', 'varest2');
N=[Link];
theta_1 = 2./N;
theta_2 = 1./N + 100./(N.^2);
figure(3);
plot(N, theta_1, N, theta_2, '.-'), legend('varest1', 'varest2');
1
2
Published with MATLAB® R2020a
3
Question-9
Code for finding MMSE if N is given
N = input('Enter the number of observations')
M = 5000
% M denotes number of realizations
mean_theta = 12;
var_theta = 0.011;
% mean_theta is prior mean %
% var_theta is prior variance %
mean_noise = 0;
var_noise = 1;
% mean_noise is noise mean %
% var_noise is noise variance %
theta_real = mean_theta + sqrt(var_theta).*randn(1,M);
noise_real = mean_noise + sqrt(var_noise).*randn(N,M);
% theta_real is M realizations of theta %
% noise_real is M realizations of random vector of noise %
meas = ones(N,M);
for j=1:M
meas(:, j) = theta_real(j) + noise_real(:,j);
end
% meas is set of measurements %
Z_bar = mean(meas, 1);
% Z_bar is sample mean %
ratio = var_theta/(var_theta+(var_noise/N));
theta_estimate = ratio*Z_bar + (1-ratio)*mean_theta;
% theta_estimate is MMSE estimate of theta %
MSE = immse(theta_real, theta_estimate)
% MSE is mean squared error %
Error using input
Cannot call INPUT from EVALC.
Error in ell_705_M (line 4)
N = input('Enter the number of observations')
Published with MATLAB® R2020a
1
Code for plotting MMSE vs no. of measure-
ments
% No. of measurements i.e. N is varied from 1 to 1000
M = 5000
% M denotes number of realizations
MSE_arr = ones(1000, 1);
for N=1:1000
mean_theta = 12;
var_theta = 0.011;
% mean_theta is prior mean %
% var_theta is prior variance %
mean_noise = 0;
var_noise = 1;
% mean_noise is noise mean %
% var_noise is noise variance %
theta_real = mean_theta + sqrt(var_theta).*randn(1,M);
noise_real = mean_noise + sqrt(var_noise).*randn(N,M);
% theta_real is M realizations of theta %
% noise_real is M realizations of random vector of noise %
meas = ones(N,M);
for j=1:M
meas(:, j) = theta_real(j) + noise_real(:,j);
end
% meas is set of measurements %
Z_bar = mean(meas, 1);
% Z_bar is sample mean %
ratio = var_theta/(var_theta+(var_noise/N));
theta_estimate = ratio*Z_bar + (1-ratio)*mean_theta;
% theta_estimate is MMSE estimate of theta %
MSE = immse(theta_real, theta_estimate);
% MSE is mean squared error %
MSE_arr(N) = MSE;
end
range = 1:1000;
plot(range, MSE_arr, '.-'), legend ('Mean Squared Error')
M =
1
5000
Published with MATLAB® R2020a
