Skip to main content
. 2023 Mar 3;13:3595. doi: 10.1038/s41598-023-30480-8

Table 5.

Ablation studies of the proposed model on the test sets of the considered datasets*.

Datasets Models Loss AS F1 TP TT IT
CIFAR-10 CMSFL-Net (4) 1.56 ± 0.06 0.73 ± 0.04 0.71 ± 0.03 0.56 87.33 6.87
CMSFL-Net (6) 1.15 ± 0.01 0.79 ± 0.02 0.77 ± 0.02 0.86 93.21 32.35
CMSFL-Net (8) 1.13 ± 0.02 0.80 ± 0.01 0.78 ± 0.01 1.17 98.11 35.01
CMSFL-Net (10) 1.20 ± 0.01 0.78 ± 0.02 0.75 ± 0.01 1.49 105.10 37.12
STL-10 CMSFL-Net (4) 1.43 ± 0.01 0.69 ± 0.01 0.66 ± 0.01 0.56 11.23 9.81
CMSFL-Net (6) 1.14 ± 0.02 0.74 ± 0.01 0.72 ± 0.10 0.86 14.11 10.29
CMSFL-Net (8) 1.10 ± 0.01 0.75 ± 0.01 0.72 ± 0.01 1.17 16.99 10.63
CMSFL-Net (10) 1.21 ± 0.01 0.72 ± 0.01 0.70 ± 0.01 1.49 18.21 11.32
ImageNet-100 CMSFL-Net (4) 1.12 ± 0.02 0.79 ± 0.01 0.74 ± 0.01 0.56 1,987.16 109.18
CMSFL-Net (6) 0.85 ± 0.01 0.86 ± 0.01 0.80 ± 0.01 0.86 2,366.42 134.49
CMSFL-Net (8) 0.83 ± 0.02 0.86 ± 0.02 0.81 ± 0.01 1.17 2,719.61 148.18
CMSFL-Net (10) 0.91 ± 0.01 0.83 ± 0.01 0.79 ± 0.01 1.49 3,101.83 162.19
COVID-CT CMSFL-Net (4) 0.87 ± 0.02 0.76 ± 0.01 0.72 ± 0.01 0.56 69.09 11.20
CMSFL-Net (6) 0.48 ± 0.01 0.83 ± 0.01 0.80 ± 0.01 0.86 80.92 13.56
CMSFL-Net (8) 0.53 ± 0.01 0.84 ± 0.01 0.81 ± 0.01 1.17 91.96 17.18
CMSFL-Net (10) 0.59 ± 0.02 0.83 ± 0.02 0.80 ± 0.02 1.49 100.81 23.05
BreakHis CMSFL-Net (4) 1.51 ± 0.01 0.70 ± 0.02 0.68 ± 0.02 0.56 198.91 7.37
CMSFL-Net (6) 1.17 ± 0.01 0.77 ± 0.01 0.73 ± 0.10 0.86 235.22 9.45
CMSFL-Net (8) 1.21 ± 0.01 0.76 ± 0.01 0.73 ± 0.01 1.17 299.12 15.12
CMSFL-Net (10) 1.25 ± 0.01 0.74 ± 0.01 0.71 ± 0.01 1.49 318.41 19.11
Br35H CMSFL-Net (4) 0.08 ± 0.01 0.97 ± 0.01 0.94 ± 0.01 0.56 67.91 4.63
CMSFL-Net (6) 0.04 ± 0.01 0.99 ± 0.01 0.99 ± 0.01 0.86 78.95 5.57
CMSFL-Net (8) 0.05 ± 0.01 0.99 ± 0.01 0.99 ± 0.01 1.17 91.72 7.10
CMSFL-Net (10) 0.07 ± 0.01 0.98 ± 0.01 0.97 ± 0.01 1.49 102.98 9.61

Lower loss and higher AS and F1 scores correspond to better performance of a model. TP, TT, and IT correspond to trainable parameters (millions), average training time per epoch (seconds), and inference time (seconds), respectively. *This information is based on experiments using 32 GB NVIDIA Tesla V100-SXM2 GPU.

Significant values are in [italics].