Obtaining bayesian error rate

3 visualizaciones (últimos 30 días)
Sowmya MR
Sowmya MR el 25 de Sept. de 2016
Editada: Massimo Zanetti el 26 de Sept. de 2016
I am new to machine learning and i have derived general expressions for bayes decision boundary and trying to plot the graph for mu=1 and sigma^2=2; Trying to plot a figure which contains both class conditional pdfs p(x|ωi) and posterior probabilities p(ωi|x) with the location of the optimal decision region. I should also obtain the bayes error rate for it. Can someone please help me? This is what i have tried so far:
%Plott Class-Conditional
fplot( @(x) ( (1/sqrt(2*pi)) * exp(- (x.^2)/2) ) ,'Linewidth',2);
hold on;
fplot( @(x) ( (1/(2* sqrt(pi))) * exp( - (x.^2 - 2*x +1)/4) ) ,'Linewidth',2);
hold off;
legend({'P(x|w1)','P(x|w2)'}, 'FontSize',14)
xlabel('x' ,'FontSize', 12)
ylabel('p(x|wi)' , 'FontSize', 12)
title('Graph of class conditional pdfs p(x|wi)', 'FontSize', 14)
%Plott Posterior
fplot(@(x) ( ((1/sqrt(2*pi)) * exp(- (x.^2)/2))/( ((1/sqrt(2*pi)) * exp(- (x.^2)/2)) + ((1/(2* sqrt(pi))) * exp( - (x.^2 - 2*x +1)/4)) ) ) )
hold on;
fplot(@(x) ( ((1/(2* sqrt(pi))) * exp( - (x.^2 - 2*x +1)/4)) / ( ((1/(2* sqrt(pi))) * exp( - (x.^2 - 2*x +1)/4)) + (( (1/sqrt(2*pi)) * exp(- (x.^2)/2)) ) )))
hold off;
  1 comentario
Massimo Zanetti
Massimo Zanetti el 26 de Sept. de 2016
Editada: Massimo Zanetti el 26 de Sept. de 2016
I am trying to get. It seems you have two Gaussian distributions, one is given by parameters MU=1 and SIGMA^2=2. What about the other one?
The optimal decision is given by the point where the two curves intersect, and error rate correrponds to the overlapping areas of the two sub-graphs. Please, give me the parameters of the other Gaussian distribution.

Iniciar sesión para comentar.

Respuestas (0)

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by