Comparison of performance of multilayer extreme learning machines in regression problems

Artificial intelligence is widely used in many fields with today's developments. Artificial neural networks (ANN) are effective in analyzing large data sets and contribute to the creation of smarter systems by saving time and effort. Extreme learning machine (ELM) allows ANNs to learn faster by simplifying the training process. It is specifically designed to overcome the challenges that arise when working on large data sets. In 2022, Kale and Karakuzu made two important developments on multi-layer extreme learning machines. By examining in detail the modeling performance of these networks in dynamic systems, researchers have determined that the proposed systems have superior modeling capabilities. The findings clearly show that multilayer extreme learning machines are effective in system modeling applications. In this thesis study, the success of these two newly developed ELM structures on regression problems was examined. The findings show that the newly developed models are effective in solving regression problems and provide more successful results than the original network structure. However, it has been observed that it is not advantageous in all cases in terms of processing speed. The results obtained indicate that the models have a wide application potential in regression analysis and can be a valuable tool in solving industrial/scientific problems.