A Normative Approach to Privacy-Preserving Recommender Systems: Integrating Matrix Factorization and Genetic Algorithms

Algorithm 1

Improved privacy genetic algorithm (APrivGene).

	Input: D, the set of binary groups or . f, the objective function or ;
	Output: a vector of hidden factors ;
	Control parameters in the initialization algorithm: set the number of hidden factors d, the privacy budget ε, the variation step η, the decay factor β < 1, the maximum number of iterations A, and the size l of the candidate solution set Ω;
	The initial candidate solution set Ω is generated randomly;
	For a = 1 to A-1 do
	Compute for each ;
	select individuals using the augmented index mechanism;
	Set Ω to null.
	For z = 1 to d do
	i = C(0,1)//draw random noise according to the standard Corsi distribution;
	positive directional variation;
	negative directional variation;
	End for η = ηβ.
	End for
	Compute for each ;
	Return