function?[mu?sigma]?=?Bayesian_parameter_est（train_patterns?train_targets?sigma）

%?Estimate?the?mean?using?the?Bayesian?parameter?estimation?for?Gaussian?mixture?algorithm估計平均值，用貝葉斯參數的估計作為簡縮高斯函數集算法
%?Inputs:
%?	patterns	-?Train?patterns
%	targets	-?Train?targets
%	sigma		-?The?covariance?matrix?for?each?class?每一類矩陣的協方差
%
%?Outputs
%	mu			-?The?estimated?mean?平均值
%	sigma		-?The?estimated?covariances?協方差

[NM]		=?size（train_patterns）;
Uclasses?=?unique（train_targets）;
Nuc		=?length（Uclasses）;

%Find?initial?estimates?for?mu?and?sigma?for?the?classes
mu0		=?zeros（Nuc?N）;
sigma0	=?zeros（Nuc?N?N）;
for?i?=?1:Nuc
	indices	=?find（train_targets?==?Uclasses（i））;
???mu0（i:）?=?mean（train_patterns（:indices）‘）;
???sigma0（i::）?=?sqrtm（cov（train_patterns（:indices）‘1））;%方根矩陣???協方差：COV（X，Y），即COV（X，Y）=E[（X-E（X））（Y-E（Y））