Table 3 CircCNN with BN outperforms other modified CircCNN

From: Prediction of Back-splicing sites for CircRNA formation based on convolutional neural networks

		AUC	ACC	MCC	Sens	Spec
Human	CircCNN (No BN)	0.8963	0.8317	0.6613	0.8871	0.766
	CircCNN (BN → Dropout)	0.8979	0.8325	0.6632	0.8891	0.7655
	CircCNN	0.9049	0.8421	0.6849	0.9147	0.7562
Mouse	CircCNN (No BN)	0.8401	0.7629	0.5275	0.8003	0.7255
	CircCNN (BN → Dropout)	0.8407	0.7623	0.5263	0.8001	0.7246
	CircCNN	0.8514	0.7705	0.5508	0.8614	0.6797
Fruit Fly	CircCNN (No BN)	0.858	0.773	0.5483	0.8058	0.7402
	CircCNN (BN → Dropout)	0.86	0.7753	0.5527	0.8165	0.7341
	CircCNN	0.8708	0.7869	0.5773	0.8374	0.7365

Each number represents the average metric value of model in cross-validation

Back to article page