Table 3 Clustering performance evaluation using the 193 raw global features
From: Deep clustering of small molecules at large-scale via variational autoencoder embedding and K-means
Clustering method | #Clusters | Internal indices | ||
|---|---|---|---|---|
Calinski–Harabasz | Silhouette | Davies–Bouldin | ||
K-means (193 molecular features) | 60 | 749.187 | 0.068 | 1.888 |
BIRCH (193 molecular features) | 60 | 706.192 | 0.059 | 1.710 |
VAE (16) + K-means | 95 | 4985.544 | 0.236 | 1.141 |
VAE (32) + K-means | 55 | 5168.007 | 0.223 | 1.160 |
VAE (64) + K-means | 65 | 4991.844 | 0.227 | 1.130 |
AE (16) + K-means | 45 | 878.878 | 0.073 | 2.009 |
AE (32) + K-means | 45 | 1112.495 | 0.090 | 1.923 |
AE (64) + K-means | 45 | 1700.526 | 0.117 | 1.688 |