Machine learning is going to become the driving technique of all most all modern technologies which will be more intelligent and efficient. Extreme Learning Machine is a computationally efficient neural network model with non-iterative learning strategy. The improved versions of ELM that are evolved within a short period of time show the enthusiasm imparted to the researchers and scientists in Computer Science especially in Artificial Intelligence. This paper intended to give a study on four important variants of ELM - Twin-ELM, HOG-ELM, ER2- ELM and RF-ELM. The MNIST and USPS datasets are used in the experiments and the results give more insight into the ELM. The performance of these four variants in terms of its computational constraints, training time and accuracy, helps to understand the direction of developments in ELM.
Volume 11 | 06-Special Issue
Pages: 826-835