An improved extreme learning machine for classification problem based on affinity propagation clustering

Extreme learning machine (ELM) is an efficient algorithm for single-hidden layer feedforward neural networks (SLFNs), which can produce good generalization performance in most cases and learn thousands of times faster than conventional popular algorithms. However, the performance of ELM is sensitive to the initialized number of hidden neurons. In some traditional methods, the number of hidden nodes is gradually increased by a fixed interval to select the nearly optimal number of nodes for ELM, whereas these methods are of a little bit of complexity and quite time-consuming. This paper proposes an improved ELM based on the affinity propagation clustering, which does not need to define the number of hidden nodes in advance manually and randomly. The proposed algorithm automatically determines the number of hidden nodes for different data sets. Empirical study of AP-based ELM on several commonly used classification benchmark problems shows that it achieves better performance compared with the standard ELM.