Table 4.

The performance of an inception-based capsule neural network on variant generalization tests ranges from 10 to 50%.

Split ratio	Accuracy	Sensitivity	Specificity	AUC-ROC
90% vs 10%	0.9928	0.9898	0.9899	0.9943
80% vs 20%	0.9935	0.9957	0.9912	0.9973
70% vs 30%	0.9894	0.9847	0.9890	0.9894
60% vs 40%	0.9849	0.9837	0.9912	0.9950
50% vs 50%	0.9810	0.9805	0.9858	0.9709