Table 3 The number of parameters and memory allocation. MoMS-Net requires the same memory allocation order as GCN. However, MassFormer requires a larger amount of memory despite having a much smaller batch size.

From: Mass spectra prediction with structural motif-based graph neural networks

 

# of parameters

Memory allocation(MB)

Batch size

CNN

7.60E+06

718

1024

MassFormer

6.61E+06

1923

128

WLN

5.25E+06

1531

1024

GCN

1.01E+07

1002

1024

MoMS-Net (Ours)

1.81E+07

1409

1024

Back to article page