Table II:

comparison of AUC scores between DGCs and the corresponding baselines on ChestX-ray14 dataset. The better results based on the same encoder are bolded. The best results among all the models are colored red.

Diseases	AN	G-AN	RN50	G-RN50	DN201	G-DN201	DN121	G-DN121	DN161	G-DN161	VN16	G-VN16
Atelectasis	0.7327	0.7359	0.7407	0.7418	0.7426	0.7415	0.7391	0.7378	0.7433	0.7421	0.7488	0.7495
Cardiomegaly	0.8681	0.8699	0.8553	0.8623	0.8541	0.8606	0.8483	0.8517	0.8662	0.8653	0.8700	0.8687
Effusion	0.7874	0.7896	0.7984	0.7991	0.7987	0.7997	0.7982	0.7976	0.8026	0.8052	0.8075	0.8096
Infiltration	0.6725	0.6771	0.6798	0.6773	0.6756	0.6790	0.6775	0.6792	0.6751	0.6773	0.6894	0.6869
Mass	0.7340	0.7374	0.7480	0.7502	0.7567	0.7614	0.7495	0.7538	0.7628	0.7652	0.7756	0.7820
Nodule	0.6778	0.6795	0.6962	0.6934	0.7107	0.7118	0.7032	0.7029	0.7142	0.7188	0.7248	0.7255
Pneumonia	0.6632	0.6744	0.6877	0.6868	0.6861	0.6917	0.6878	0.6858	0.6939	0.7020	0.6943	0.6954
Pneumothorax	0.8149	0.8195	0.8364	0.8405	0.8371	0.8396	0.8356	0.8405	0.8437	0.8497	0.8441	0.8451
Consolidation	0.7088	0.7105	0.7112	0.7136	0.7129	0.7177	0.7124	0.7167	0.7174	0.7190	0.7259	0.7283
Edema	0.8280	0.8311	0.8273	0.8311	0.8294	0.8302	0.8287	0.8280	0.8355	0.8368	0.8363	0.8340
Emphysema	0.7991	0.8059	0.8659	0.8724	0.8704	0.8668	0.8579	0.8590	0.8793	0.8823	0.8707	0.8699
Fibrosis	0.7956	0.7977	0.7939	0.7885	0.7982	0.8036	0.7873	0.7905	0.7972	0.8103	0.7979	0.7978
PT*	0.7356	0.7373	0.7379	0.7424	0.7424	0.7432	0.7465	0.7457	0.7451	0.7510	0.7554	0.7581
Hernia	0.8485	0.8498	0.8879	0.8816	0.8974	0.9113	0.8951	0.8898	0.9097	0.9015	0.8837	0.8776
Average	0.7619	0.7654	0.7762	0.7772	0.7794	0.7827	0.7762	0.7771	0.7847	0.7876	0.7875	0.7877

^*PT — Pleural Thickening; AN — AlexNet; VN16 — VGGNet16; RN50 — ResNet50; DN121 — DenseNet121; DN161 — DenseNet161; DN201— DenseNet201; G-AN — DGC-AlexNet; G-VN16 — DGC-VGGNet16; G-RN50 — DGC-ResNet50; G-DN121 — DGC-DenseNet121; G-DN161 — DGC-DenseNet161; G-DN201 — DGC-DenseNet201.