基于区间边界探测的进化随机配置网络  

Boundary-detection-based evolutionary stochastic configuration networks

作　　者：吉建娇[1] 王殿辉 JI Jian-jiao;WANG Dian-hui(Artificial Intelligence Research Institute,China University of Mining and Technology,Xuzhou Jiangsu 221116,China;State Key Laboratory of Synthetical Automation for Process Industries,Northeastern University,Shenyang Liaoning 110819,China)

机构地区：[1]中国矿业大学人工智能研究院,江苏徐州221116 [2]东北大学流程工业综合自动化国家重点实验室,辽宁沈阳110819

出　　处：《控制理论与应用》2024年第10期1913-1922,共10页Control Theory & Applications

基　　金：国家重点研发计划项目(2018AAA0100304);国家自然科学基金项目(62203444)资助.

摘　　要：随机配置网络在监督机制下利用随机采样方式在可调区间内分配新增隐节点参数,易陷入局部最优,导致模型包含不必要的冗余节点.为保证模型的紧致性,本文提出了一种进化随机配置网络架构,利用进化算法迭代搜索隐节点参数的最优配置.首先,初始化策略通过探测有潜力的区间边界生成满足监督机制的初始种群,以加快种群的收敛速度.随后,采用基于Q-learning的选择策略自适应地根据种群所在区间边界为进化算子选取合适的参数设置,进而提高其搜索效率,增强种群的收敛性能.最后,在基准数据集上的实验表明了所提算法构建模型的紧致性和精度.The stochastic configuration network introduces a supervisory mechanism to assign the parameters of newlyadded 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 population 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 efficiency,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.

关 键 词：随机配置网络 进化优化 强化学习 Q-LEARNING 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]