B	reference-dose projection domain
A	low-dose projection domain
a	training samples ${\left\{a_i\right\}}_{i=1}^Q$ where $a\in A$
b	training samples ${\left\{b_j\right\}}_{j=1}^P$ where $b\in B$
$a~p(a)$$b~p(b)$ data distribution
$AB$	mapping  $A\to B$
$BA$	mapping  $B\to A$
DA	distinguish between images $\left\{a\right\}$
	and translated images $BA\left\{b\right\}$
DB	discriminate between $\left\{b\right\}$nd $AB\left\{a\right\}$
D	discriminator
G	generator