I have a linear map, $J$, from one space to another. I can transform a covariance matrix, $P$, from one space to another by using:

$P' = JPJ^{T}$

However, in my situation I have an inverse covariance matrix, $P^{-1}$

The simplest way that I can think of transforming this is:

$P'^{-1} = (J(P^{-1})^{-1}J^{T})^{-1}$

which then using

$(A*B*C)^{-1} = C^{-1}B^{-1}A^{-1}$

should give

$P'^{-1} = (J^{T})^{-1}P^{-1}J^{-1}$

however, when I try this in MATLAB I do not find that these two representations are equal.

$(J(P^{-1})^{-1}J^{T})^{-1} = (J^{T})^{-1}P^{-1}J^{-1}$


It is actually correct.

I was comparing the two matrices using "==" which has no epsilon for comparing floating point numbers.

$(J(P^{−1})^{−1}J^{T})^{−1} = (J^{T})^{−1}P^{−1}J^{−1} = (JPJ^{T})^{-1}$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.