An efficient annealing-assisted differential evolution for multi-parameter adaptive latent factor analysis

Journal of Big Data

Table 1 Classification of parameter adaptive mechanisms

Classifications	Adaptive mechanisms	Defects	HDI data analysis	Multi-parameters
Analytical approach	Simulated annealing approach[31]	Internal assumptions	×	×
Analytical approach	Parameter estimation adjustment[32]	Manual tunning	×	×
Dynamic optimization	NE[33]	Need Priori Knowledge	×	×
Dynamic optimization	Gradient-based estimation[34]	Sensitive to initialization	×	×
Evolutionary computation	PSO[36]	Premature convergence	√	×
Evolutionary computation	BSO[37]	time consuming	√	√

ADMA employs evolutionary computation DE as the multi-parameter adaptive mechanism and improve its optimization progress by an annealing technique. Therefore, ADMA inherits the flexibility and adaptively of swarm intelligence, while enjoying the exploration and exploitation merits of the improved DE, which thereby can solve the big data analysis problems well especially conducting an LFA on an HDI matrix