Fig. 5: Visualizing the multi-objective RL composition space.
From: Deep reinforcement learning for inverse inorganic materials design
Most commonly found elements in compounds generated by a PGN and b DQN models with w ∈ {0.2,0.4,0.6,0.8} spanning the synthesis-property space. Elements are color-coded by their identity. Dark blue squares lacking a label correspond to not enough data to plot.
