Skip to main content
. 2017 Sep 4;9:48. doi: 10.1186/s13321-017-0235-x

Fig. 6.

Fig. 6

Average similarity J of generated structures as a function of training steps. Difference in learning dynamics for the Agents based on the canonical Prior, and those based on a reduced Prior where everything more similar than J=0.5 to Celecoxib have been removed