Skip to main content
. 2019 Dec 14;19(24):5541. doi: 10.3390/s19245541

Table 5.

Accuracy of the best candidate models found using multi-filtering search, then trained and tested on CIFAR-10. The first group contains networks with larger configurations, while the second group comprises smaller ones.

Model Widening Factor k Mean Acc Mil. FLOPs
IoTNet-3-2 1.08 89.79% 11
IoTNet-3-4 0.7 89.9% 9.9
IoTNet-3-3 0.66 88.98% 6.2
IoTNet-3-2 0.68 87.19% 4.2
IoTNet-3-2 0.5 81.47% 2.6
IoTNet-3-3 0.41 83.49% 2.5