Sparse extreme learning machine for classification

Extreme Learning Machine (ELM) is a single-hidden-layer feed forward neural network that randomly initiates the weights for the connections between input and hidden layer and the bias of the hidden layer. Two hidden layers and multiple hidden layers ELMs have added some enhancements – when used on the same data set - keeping the randomly initiated weights and biases and extending the number of hidden layers. A greedy algorithm is also proposed within the ELM architecture to penetrate sparse approaches and presented enhancement in compare with standard. This study presents three different standard ELM algorithms and a greedy single hidden layer ELM to detect classification problems on 3 datasets and compares the testing accuracy of these total 4 ELM algorithms. Based on the results of experiments conducted, greedy Extreme Learning Machine algorithm provide a good prediction accuracy with a high prediction rate.