A Clonal Selection Optimization System for Multiparty Secure Computing

Algorithm 3

Clone selective optimized gradient descent (CSGD).

	Input:
	Objective function , loss function , and affinity value is
	Initial:
	Some sample datasets for features , is sample features, are model parameters
	For model : all , stride size , mutation probability is , clone probability is , ending distance =
	Determine the gradient corresponding to , the current position, the gradient is
	For each iteration
	For all random sample data do
	Generate the gradient vector and the corresponding value of during model learning
	Calculate according to
	Reserve the gradient vector group with bigger
	Optimal selection of gradient vector by clonal selection strategy:
	Calculate according to , population adjust as
	According to , population adjusted as
	Retain to original population size:
	Keep excellent vectors:
	Iterative condition judgment
	If
	Output:
	Update the population based on the selected antibodies