Table 4.

Test error (%) on the smallNORB classification task.

Method	Reconstruction	Parameters (K)	smallNORB (%)
Our baseline	No	198	5.9
Base-CapsNet	No	167	4.58
Our baseline	Yes	198	4.59
Base-CapsNet	Yes	167	4.33
Efficient-CapsNet	Yes	151	$2.{54}_{\pm 0.003}$
Baseline14	No	4200	5.2
Matrix-CapsNet with EM routing14	No	310	1.8 ($4.4_{\pm 0.004}$)*
CapsNet10	Yes	6800	3.77
VB-Routig16	Yes	310	$1.6_{\pm 0.06}$

All methodologies are reported with their number of parameters andthe presence of the reconstruction regularizer during the training phase. * indicates the results from our experiments.