BAYESDF |
Computes decision boundary of Bayesian classifier.
Synopsis:
quad_model = bayesdf(model)
Description:
This function computes parameters of decision boundary
of the Bayesian classifier with the following assumptions:
- 1/0 loss function (risk = expectation of misclassification).
- Binary classification.
- Class conditional probabilities are multivariate Gaussians.
In this case the Bayesian classifier has the quadratic
discriminant function
f(x) = x'*A*x + B'*x + C,
where the classification strategy is
q(x) = 1 if f(x) >= 0,
= 2 if f(x) < 0.
Input:
model [struct] Two multi-variate Gaussians:
.Mean [dim x 2] Mean values.
.Cov [dim x dim x 2] Covariances.
.Prior [1x2] A priory probabilities.
Output:
quad_model.A [dim x dim] Quadratic term.
quad_model.B [dim x 1] Linear term.
quad_model.C [1x1] Bias.
Example:
trn = load('riply_trn');
tst = load('riply_trn');
gauss_model = mlcgmm(trn);
quad_model = bayesdf(gauss_model);
ypred = quadclass(tst.X,quad_model);
cerror(ypred,tst.y)
figure; ppatterns(trn); pboundary(quad_model);
See also
BAYESCLS, QUADCLASS
About: Statistical Pattern Recognition Toolbox
(C) 1999-2003, Written by Vojtech Franc and Vaclav Hlavac
Czech Technical University Prague
Faculty of Electrical Engineering
Center for Machine Perception
Modifications:
18-oct-2005, VF, dealing with Cov given as vector repared
01-may-2004, VF
19-sep-2003, VF
24. 6.00 V. Hlavac, comments into English.