Workflow from Scientific Research

High-level diagram displaying the architecture of the deep reinforcement learning model used by Popova and colleagues 40 and Olivecrona and colleagues 38 . Left: supervised learning language model. The prior model learns to generate novel SMILES from a large dataset of SMILES from ChEMBL. Right: reinforcement learning. determine the effect that training data representationThe agent model (based on the prior model) is trained to generate SMILES that return a favourable reward.

DOI