Fig. 2From: GEN: highly efficient SMILES explorer using autodidactic generative examination networksModeling workflow used for every architecture/hyperparameter search. The autodidactic generator models learn independently a probability for the next logical character. At every epoch, i.e. online the generator generates a statistical sample of 300 SMILES Strings with are examined using statistical quality control as examination criteria. Upon completion of the training, the earliest stable model that satisfies the quality criteria is selected and evaluated based on a generated sampleBack to article page