Extreme Learning Regression for nu Regularization
Extreme learning machine for regression (ELR), though efficient, is not preferred in time-limited applications, due to the model selection time being large. To overcome this problem, we reformulate ELR to take a new regularization parameter nu (nu-ELR) which is inspired by Schölkopf et al. The regul...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Taylor & Francis Group
2020-04-01
|
Series: | Applied Artificial Intelligence |
Online Access: | http://dx.doi.org/10.1080/08839514.2020.1723863 |
Summary: | Extreme learning machine for regression (ELR), though efficient, is not preferred in time-limited applications, due to the model selection time being large. To overcome this problem, we reformulate ELR to take a new regularization parameter nu (nu-ELR) which is inspired by Schölkopf et al. The regularization in terms of nu is bounded between 0 and 1, and is easier to interpret compared to C. In this paper, we propose using the active set algorithm to solve the quadratic programming optimization problem of nu-ELR. Experimental results on real regression problems show that nu-ELR performs better than ELM, ELR, and nu-SVR, and is computationally efficient compared to other iterative learning models. Additionally, the model selection time of nu-ELR can be significantly shortened. |
---|---|
ISSN: | 0883-9514 1087-6545 |