Skip to main content

Table 4 The average MSE of predicting randomly distributed test values using different auto-encoder models

From: A Graph Feature Auto-Encoder for the prediction of unobserved node features on biological networks

Model E. coli Mus Musclus
TF_Net PPI Genetics PPI
GCN 0.043 ± 0.00175 0.065 ± 0.004 0.114 ± 0.004 0.011 ± 0.001
GraphSAGE 0.027 ± 0.0007 0.023 ± 0.0004 0.026 ± 0.0003 0.004 ± 0.0006
GraphConv 0.041 ± 0.003 0.068 ± 0.05 0.182 ± 0.046 2.06 ± 2.73
FeatGraphConv (our) 0.025 ± 0.0008 0.023 ± 0.0006 0.025 ± 0.0004 0.003 ± 0.0002
MLP Auto-encoder 0.031 ± 0.0007 0.028 ± 0.0003 0.027 ± 0.0004 0.004 ± 0.0005
MAGIC 3.505 ± 0.006 3.661 ± 0.017 3.215 ± 0.003 0.050 ± 0.0002
  1. (Bold indicates lowest error mean per network)