Scribd
Upload
152 views1 page

Gaussian Identities

The document discusses important identities involving Gaussian distributions. It states that: 1) The marginals and conditionals of a multivariate Gaussian are also Gaussian distributions. 2) If one variable has a Gaussian distribution, and another is conditioned on the first with an additional Gaussian, their joint distribution is Gaussian. 3) In this case, the marginal of the second variable is Gaussian, and the conditional of the first given the second is also Gaussian.

Uploaded by

Bob
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
152 views1 page

Gaussian Identities

The document discusses important identities involving Gaussian distributions. It states that: 1) The marginals and conditionals of a multivariate Gaussian are also Gaussian distributions. 2) If one variable has a Gaussian distribution, and another is conditioned on the first with an additional Gaussian, their joint distribution is Gaussian. 3) In this case, the marginal of the second variable is Gaussian, and the conditional of the first given the second is also Gaussian.

Uploaded by

Bob
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

Gaussian Identities

Michael A. Osborne
[email protected]
January 27, 2014

The following identities are almost indispensible when dealing with Gaussian dis-
tributions, which we denote in the usual way as
 
1 1 T 1
N (x; , A) , exp (x ) A (x ) .
det 2A 2

Probably the most important properties of a multivariate Gaussian distribution are


that its marginals and conditionals are both themselves Gaussian. That is, if we have
     
x A C
p( x, y | I ) , N ; , ,
y CT B

then its marginal and conditional distributions are respectively

p( x | I ) = N (x; , A) (1)
1 1 T

p( x | y, I ) = N x; + C B (y ), A C B C . (2)

We now turn to another important property of the Gaussian distribution. If we


have

p( x | I ) , N (x; , A)
(3)
p( y | x, I ) , N (y; M x + c, L) ,

then the joint distribution can be written as

A MT
     
x A
p( x, y | I ) = N ; , , (4)
y M + c MA L + M A MT

and so, using (1) and (2)

p( y | I ) = N y; M + c, L + M A MT

(5)
p( x | y, I ) = N (x; + (y M c), A M A) (6)

where 1
= A MT L + M A MT

You might also like