Propose a version of the subspace identity model that is robust to outliers in the training data. Moghaddam et al XXXXXXXXXXtook a different probabilistic approach to face verification. They took the...

Propose a version of the subspace identity model that is robust to outliers in the training data.

Moghaddam et al. (2000) took a different probabilistic approach to face verification. They

<sup>took




the




difference





x</sup>


∆

<sup>=





x</sup>


2


<sup>−








x</sup>


1

<sup>and




modeled




the




likelihoods




of




this




vector





Pr



(



x</sup>


∆


<sup>|




w







=




0)




and





Pr



(



x</sup>


∆


<sup>|




w







=




1)




when




the




two




faces




match




or




don’t.




Propose




expressions




for




these




likelihoods




and</sup>



discuss learning and inference in this model. Identify one possible disadvantage of this model.