Conditional of Gaussian distribution
Problem
Given $(X, Y)\in N(\mu, \Lambda)$ find $Y|X=x$.
Gaussian density
Solution
Symbols used
In our case we have (by definition)
and the inverse is
From definition of conditional distribution and the definition of gaussian density \eqref{eq:gaussiandefinition} we get
General case
Conditioning of a general multivariate normal distribution (thanks to Jean Alexander for the hints).
Both $X$ and $Y$ are in general multivariate normal distributions (since $(X, Y)$ are) Where the elements are actually blocks of the matrix but it is just another representation from what we normally use to be able to represent the conditioning explicitly with these symbols.
Then we have that the general conditioning becomes (this will not be proved)
Where the operation $A^{-1}$ is the generalized inverse and satisfies the condition $AA^{-1}A = A$. Note that if the matrix is invertible then the generalized inverse is just the ordinary matrix inversion.
The source: Gaussian Processes