A study on effectiveness of extreme learning machine

An improved algorithm called EELM is proposed that makes a proper selection of the input weights and bias before calculating the output weights, which ensures the full column rank of H in theory and improves to some extend the learning rate and the robustness property of the networks.

Abstract

Extreme learning machine (ELM), proposed by Huang et al., has been shown a\npromising learning algorithm for single-hidden layer feedforward neural\nnetworks (SLFNs). Nevertheless, because of the random choice of input weights\nand biases, the ELM algorithm sometimes makes the hidden layer output matrix H\nof SLFN not full column rank, which lowers the effectiveness of ELM. This paper\ndiscusses the effectiveness of ELM and proposes an improved algorithm called\nEELM that makes a proper selection of the input weights and bias before\ncalculating the output weights, which ensures the full column rank of H in\ntheory. This improves to some extend the learning rate (testing accuracy,\nprediction accuracy, learning time) and the robustness property of the\nnetworks. The experimental results based on both the benchmark function\napproximation and real-world problems including classification and regression\napplications show the good performances of EELM.\n