# conditioned
`conditioned` method is like the mapping $\mathcal{F}$ such that
$$
\mathcal{F}: p(a \mid b) \mapsto p(a \mid f(c))
$$
with mapping $f$ such that
$$
f: C \mapsto B
$$
Log
$$
\log{p(\mathbf{a} \mid f(c))}
$$
Log Value Difference
$$
\frac{\partial \log{p(\mathbf{a} \mid f(c))}}{\partial \mathbf{a}}
$$
Log Condition Difference
（ use `ConditionDifferentiableConditionedDistribution` ）
$$
\frac{\partial \log {p(\mathbf{a} \mid f(c))}}{\partial c} = \frac{\partial \log {p(\mathbf{a} \mid f(c))}}{\partial f(c)} \times \frac{\partial f(c)}{\partial c}
$$
`ConditionDifferentiableConditionedDistribution` has
Conditioned Distribution : $ p(\mathbf{a} \mid f(c))$ ,
and
Differentiated Condition ( not $\log$ ) : $\frac{\partial f(c)}{\partial c}$ ( Matrix )