Table 5:

COVID-19 Omics data: SVM is based on stacked three-view raw data. DeepCCA+SVM is training SVM based on the last layer of DeepCCA. DeepSIDA applies nearest centroid on the last layer for classification. The reported classification accuracy for Deep IDA are based on optimized network structure. DeepIDA + top 10% is based on input-512–20 network structure. DeepIDA + top 50 is based on input-512-256-20 network structure.

Method	Accuracy (%)
SVM	49.01
SVM on selected top 10% features	62.74
SVM on selected top 50 features	64.71
Deep GCCA + SVM	64.71
Deep GCCA + SVM on selected top 10% features	68.63
Deep GCCA + SVM on selected top 50 features	64.71
SIDA	60.78
Deep IDA	76.47
Deep IDA on selected top 10% features	76.47
Deep IDA on selected top 50 features	78.43