1: \begin{definition}

2: Let the following restriction be applied to \eq{eqn:fisher_neyman} in Theorem~\ref{thm:fisher_neyman},

3: \begin{equation}

4: \label{eqn:exp_fam}

5: g( T(x) ; \vartheta) = \exp\big\{ \langle \eta(\vartheta),  T(x) \rangle - A(\vartheta) \big\},

6: \end{equation}

7: where $\langle \cdot, \cdot \rangle$ denotes the usual inner product in Euclidean space, and $\eta(\vartheta)$, $T(x)$, and $A(\vartheta)$ are known functions.  The resulting class of probability distributions spanned by \eq{eqn:fisher_neyman}, with $g(\bullet)$ given by (\ref{eqn:exp_fam}), is known as the {\em exponential family} of probability densities.

8: \end{definition}

9: