---

Software page for fast (mixture of) transformation invariant component analysis

[pc, mu, phi, lam, lpx] =mtca(x, C, K, MFA, NITER);

Uses NIT iterations of Expectation-Maximization (EM) algorithm to learn a mixture of transformation invariant component analysis model with C classes and K factors for each class using T training images, x(:,:,t); Here, MFA= 1, learns a mixture of factor analysis and MFA =0 learns MTCA.

Usage:

load data

x=x(:,:,1:3:end);

[pc, mu, phi, lam, lpx] =mtca(x, 3, 2, 0, 20);

Refer to main webpage for more details about the model and the links to the relevant paper(s).