ABSTRACT |
In this dissertation, a new constructive algorithm for automatic design of the Adaptive Topology Extreme Learning Machine (AT-ELM) topology, named the Gaussian kernel approximation (GKA) algorithm, is proposed, developed, analyzed and validated. The approach used by the algorithm, which is based on a constructive learning algorithm, is to create the hidden units directly so that automatic design of the architecture of neural networks can be carried out. The algorithm is de_ned using the linear summation of input patterns and their randomized input weights. Hidden-layer nodes are de_ned so as to partition the input space into homogeneous regions, where each region contains patterns belonging to the same class. The algorithm is tested on three benchmark data sets of di_erent dimensionality and sample sizes to compare the approach presented here with other algorithms. Real medical diagnoses and a biological classi_cation of mushrooms are used to illustrate the performance of the algorithm. These results con_rm that the algorithm proposed here can discover the optimal number of hidden nodes for each problem and provide good results for each classi_cation problem. Moreover, the experimental results also show that the new algorithm can attain consistently good classi_cation performance, as good as that of gradient-based, analytical based, and rule-based classi_ers. |