论文部分内容阅读
隐层节点数是影响极端学习机(ELM)泛化性能的关键参数,针对传统的ELM隐层节点数确定算法中优化过程复杂、容易过学习或陷入局部最优的问题,提出结构风险最小化-极端学习机(SRM-ELM)算法。通过分析VC维与隐层节点数量之间的关联,对VC信任函数进行近似改进,使其为凹函数,并结合经验风险重构近似的SRM。在此基础上,将粒子群优化的位置值直接作为ELM的隐层节点数,利用粒子群算法最小化结构风险函数获得极端学习机的隐层节点数,作为最优节点数。使用6组UCI数据和胶囊缺陷数据进行仿真验证,结果表明,该算法能获得极端学习机的最优节点数,并具有更好的泛化能力。
The number of hidden layer nodes is a key parameter that affects the generalization performance of Extreme Learning Machine (ELM). In order to solve the problem that the optimization process is complex, easy to learn or fall into the local optimum in ELM algorithm, - Extreme Learning Machine (SRM-ELM) algorithm. By analyzing the correlation between the VC dimension and the number of hidden nodes, the VC trust function is approximately improved to make it a concave function, and the approximate SRM is reconstructed with empirical risk. On this basis, the particle swarm optimization position value is taken as the hidden layer node number of the ELM, and the particle swarm algorithm is used to minimize the structural risk function to obtain the number of hidden layer nodes of the extreme learning machine as the optimal node number. Six groups of UCI data and capsule defect data were used to verify the simulation results. The results show that the algorithm can get the optimal number of nodes and achieve better generalization ability.