Algorithm 3: Constrained Molecule Generation

Input: ${\widehat{\theta }}_A$, trained adaptor parameters;                ${\widehat{\theta }}_G$, pre-trained generator parameters;                ${\theta }_D$, pre-trained decoder parameters;                z, random noise.   Output: ${\mathcal{M}}_{gen}$, generated desired drug molecules;   Hyperparameters: $N_{\mathit{gen}}\in {\mathbb{Z}}^{*}$ 1 $r_G\leftarrow {\widehat{\theta }}_G\left({\widehat{\theta }}_A\left(z\right)\right)$ 2 ${\mathcal{M}}_{gen}\leftarrow {\widehat{\theta }}_D\left(r_G\right)$ 3 return ${\mathcal{M}}_{gen}$