Further improvements on extreme learning machine for interval neural network

• Published in: Neural computing & applications Vol. 29; no. 8; pp. 311 - 318
• Main Authors: Yang, Li-fen, Liu, Chong, Long, Hao, Ashfaq, Rana Aamir Raza, He, Yu-lin
• Format: Journal Article
• Language: English
• Published: London Springer London 01.04.2018; Springer Nature B.V
• Subjects: Artificial Intelligence; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Data Mining and Knowledge Discovery; Fuzzy logic; Fuzzy sets; Image Processing and Computer Vision; Iterative methods; Learning; Neural networks; Original Article; Performance prediction; Probability and Statistics in Computer Science; Regression analysis; Interval-valued data; Interval ELM; Generalized inverse; ELM
• DOI: 10.1007/s00521-016-2727-4
• ISSN: 0941-0643

The interval extreme learning machine (IELM) (Yang et al. in Neural Comput Appl 27(1):3–8, 2016 ) is a newly proposed regression algorithm to deal with the data with interval-valued inputs and interval-valued output. In this paper, we firstly analyze the disadvantages of IELM and further point out that IELM is actually a slight variant of fuzzy regression analysis using neural networks (Ishibuchi and Tanaka in Fuzzy Sets Syst 50(3):257–265, 1992 ). Then, we propose a new interval-valued ELM (IVELM) model to handle the interval-valued data regression. IVELM does not require any iterative adjustment to network weights and thus has the extremely fast training speed. The experimental results on data sets used in (Yang et al. 2016 ) demonstrate the feasibility and effectiveness of IVELM which obtains the better predictive performance and faster learning speed than IELM.