Theorem 1(PDF Projection theorem), see [3,5].
be a mapping from
be an arbitrary feature PDF with support .
be a reference distribution with support .
, the distribution imposed on
be non-zero and have finite value everywhere on .
Then, the function (2.2)
is a PDF (integrates to 1 over ), and is a member of
Proof: These assertions are proved in Section 16.4.
See also  or , Theorem 2.