Uncertainty Estimation Using Variational Mixture of Gaussians Capsule Network for Health Image Classification

Table 1

Comparison of the error rates between the VMG-Routing model and some models from the literature. () indicates the models that our device could not implement due to memory limitations. The values reported here were thus obtained from the literature. (−) indicates unavailable values. (#c) represents the number of classes in the dataset.

Algorithm	Error rate (%)
	CIFAR-10	Fashion-MNIST	MNIST	COVID-19 radiography
VB-routing {64, 8, 16, 16, #c} [11]	13.10	5.46	0.99	8.01
VB-routing {64, 16, 32, 32, #c} [11]	11.2 .09	5.61	—	—
VB-routing {64, 16, 16, 16, #c} [11]	12.40	5.2 .07	—	—
EM-routing {64, 16, 16, 16, #c} [4]	—	6.14	—	—
EM-routing {64, 16, 32, 32, #c} [4]	11.2 .09	—	—	—
Multi-lane LBP-gabor capsule [32]	11.43	5.17	1.00	8.04
Dynamic routing [3]	35.57	22.45	0.25	8.09
VMG-routing {32, 4, 8, 8, #c} (ours)	12.19	5.38	1.00	7.01