A=BAYES(C,H) returns a bayes object initialized with density
estimator C and hyperparameters H.
Training will try to fit the estimator C to each class of
pattern recognition data. C can be an array of algorithms, one
for each class.
Testing will return the class with the highest probability
estimate, or if passed the empty dataset will generate new
class data according to the densities learnt
Hyperparameters:
l=50 -- number of data points to generate if asked to generate
train_priors=1 -- whether to adjust priors or not according to data
Model:
child=gauss -- array of underlying density estimators
prior=1 -- prior probabilities for each class, default equal
Methods:
train, test, generate
Examples:
get_mean(train(cv({bayes svm}),toy))
get_mean(train(cv({bayes(gauss('assume="diag_cov"')) svm}),toy))
gen(bayes({gauss([-1]) gauss([1])}))
d=gen(bayes({gauss([-1 3]) gauss([0 4]) gauss([1 2])}))
| |
Reference : chapter 2 (Richard O. Duda and Peter E. Hart) Bayesian Decision Theory |
| |
Author : Richard O. Duda , Peter E. Hart |