High performance logistic regression for privacy-preserving genome analysis

BMC Medical Genomics

Table 1 Accuracy results obtained with fivefold cross-validation with LR, using the traditional sigmoid activation function and cross-entropy loss

	# no regularization (%)	Ridge regression (%)	Lasso regression (%)
BC-TCGA	99.57	99.57	99.57
GSE2034	68.83	68.84	68.83

Models were trained for 100 iterations. All computations are done in-the-clear, i.e. without use of the privacy-preserving protocols proposed in this paper

ISSN: 1755-8794