Possibility of incorporating RL

[EXJUSTICE]

Reinforcement learning uses MLP architectures by design, hence in theory it should be possible to implement in OpenChem. Rewards tend to be passed into the model as a loss via LogP or QED in literature. Would such an approach be feasible in OpenChem?

(Do RL myself, but no familiarity with molecular design)