Is there a function that does Multivariate Gaussian Regression?

Andreas Aristidou

2015 11 14

0 个回答

8 次查看（30 天）

0 个投票

Hi,

I have implemented a multivariate regression model using the MATLAB mvregress function. The model has been trained using data from 108 dances (each dance is described using 31 measurements). Dances have different emotions, which can be expressed using two variables, ρ and θ. The input of the model is a dance with an unknown emotion. The MV regress model works fine, but the results are not always accurate since the model used is linear. I was looking for a function that does multivariate gaussian regression or a way to modify the mvregress function for Gaussian instead of linear regression.

Xmat's size is 108 x 31

Y's size is 108 x 2 (ρ and θ)

Input's size is 1 x 31

n = 108;

d = 2;

Xcell = cell(1,n);
for i = 1:n
    Xcell{i} = [kron([Xmat(i,:)],eye(d))];
end
[beta,sigma,E,V] = mvregress(Xcell,Y);
B = reshape(beta,2,32)';
ptb = bsxfun(@times,Input(1,:),B(2:32,1)');
ptb = sum(ptb,2);
rho = bsxfun(@plus,[B(1,1) 0]',ptb); 
ptb = bsxfun(@times,Input(1,:),B(2:32,2)'); 
ptb = sum(ptb,2);
theta = bsxfun(@plus,[B(1,2) 0]',ptb);