Ge-SAND: an explainable deep learning-driven framework for disease risk prediction by uncovering complex genetic interactions in parallel

Table 5 The performance of Ge-SAND with sample sizes of 500 to 5,000

Sample Size	AUC-ROC (Ge-SAND VS Ridge)	Training time/epoch	Peak GPU memory consumption (GB)	Batch size	Features number
500	0.664:0.584	5.86 secs(175 steps)	2.738	2	2,594
1,000	0.696:0.6364	11.46 secs(350 steps)	2.738	2	2,594
5,000	0.7087:0.6864	57.30 secs(1,750 steps)	2.738	2	2,594
10,000	0.7222:0.6984	1 mins53.6 secs(3,500 steps)	2.738	2	2,594
50,000	0.7441:0.7298	9 mins25.88 secs(17,500 steps)	2.738	2	2,594

ISSN: 1471-2164