Optimized Pruned Annular Extreme Learning Machines

Singh, L. and Chetty, G.

Data mining with big datasets and large samples can be problematic, due to increase in complexity and computational times, and bad generalization due to outliers. Using the motivation from extreme learning machines (ELM), in this paper, we propose a novel approach based on annular ELM, involving RANSAC multi model response regularization. Experimental results on different benchmark datasets showed that proposed algorithm based on annular ELM can optimally prune the hidden nodes, and allow better generalization and higher classification accuracy to be achieved as compared to other algorithms, including SVM and OP-ELM for binary and multi-class classification and regression problems.

Cite as: Singh, L. and Chetty, G. (2014). Optimized Pruned Annular Extreme Learning Machines. In Proc. Twelfth Australasian Data Mining Conference (AusDM14) Brisbane, Australia. CRPIT, 158. Li, X., Liu, L., Ong, K.L. and Zhao, Y. Eds., ACS. 101-111

(from crpit.com) (local if available)