Figure 4.

Deep reinforcement learning consists of a generator which is usually an artificial neural network and a de novo drug-design agent that uses reinforcement learning to make decisions for the generation of novel molecular structures.