Boundary-detection-based evolutionary stochastic configuration networks
JI Jian-jiao
WANG Dian-hui
Abstract：The stochastic configuration network introduces a supervisory mechanism to assign the parameters of newly-added hidden node in the adjustable interval.However,the parameters randomly generated are easily falling into local optima,resulting in the redundant nodes embedded into the model.In order to obtain the compact model,a framework of evolutionary stochastic configuration network is proposed,the optimal parameters of hidden node are iteratively searched through an evolutionary algorithm.Firstly,the evolutionary algorithm utilizes an initial strategy to produce initial popula-tion satisfying the supervisory mechanism by detecting the promising interval boundary,with the purpose of speeding up the convergence.Following that,it employs a Q-learning-based selection strategy to automatically choose the appropriate parameters for evolution operators according to the interval boundary of population,so as to improve their searching effi-ciency,and enhance the convergence of population.Finally,the experimental results on the benchmark data demonstrate that the superior performance of the model constructed by proposed algorithm in terms of compactness and accuracy.
Keywords：stochastic configuration networksevolutionary optimizationreinforcement learningQ-learning
Publication Date：2024-10-28
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：10( 1913-1922 )