Interpretable CNN for ischemic stroke subtype classification with active model adaptation

Table 8 Comparison of different loss function for our model

Loss function	Accuracy	AUC	Recall	Precision	F1-score
Mean absolute error	0.4647	0.5100	0.4647	0.3886	0.3115
Mean absolute percentage error	0.4933	0.5082	0.4933	0.4457	0.3383
Mean squared error	0.5189	0.5464	0.5189	0.5085	0.3908
Mean squared logarithmic error	0.5643	0.5928	0.5643	0.5895	0.4693
Categorical Cross entropy	0.5665	0.6515	0.5665	0.5940	0.4863
Kullback leibler divergence	0.5660	0.6532	0.5660	0.5939	0.4815
Focal loss	0.2287	0.6104	0.2287	0.5704	0.2379
KL-focal loss	0.6020	0.6757	0.6020	0.6213	0.5141

ISSN: 1472-6947