In recent years, the interest in the study of outlier robustness properties in Extreme Learning Machines (ELM) has grown. Most of the published works uses a more robust estimation method than the commonly adopted ordinary least squares. Nevertheless, the ELM network offers other challenges that also influence its robustness properties, such as the number of hidden neurons and the computational stability of the hidden layer’s output matrix. That being said, we propose here two networks: ROP-ELM and ROPP-ELM that address the three aforementioned problems at once, in a combination of a pruning method, a cost function based on ℓ1-norm and the addition of a biologically plausible mechanism named Intrinsic Plasticity.
CITATION STYLE
Freire, A. L., & Neto, A. R. R. (2017). A robust and optimally pruned extreme learning machine. In Advances in Intelligent Systems and Computing (Vol. 557, pp. 88–98). Springer Verlag. https://doi.org/10.1007/978-3-319-53480-0_9
Mendeley helps you to discover research relevant for your work.