Table 1 Clustering of real compounds (public and in-house dataset) by k-means (k = 10) and classification into α, β, and γ. If the number of α is not zero and that of γ is more than that of β, this seemed to be predictive of classification accuracy for the generative model (light grey column). On the other hand, if the number of α is not zero and that of γ is less than that of β, this seemed to be unpredictive (dark grey column)
