208
Extreme Learning Machine: A new approach for genomic prediction of complex traits
Nowadays, a wide range of methods for predicting phenotypes based on genomic data has become available. Increasingly, the focus is also set to machine learning methodology. Despite this progress, the prediction of complex traits from high density SNP-panels remains an extremely demanding task particularly in terms of the computational effort required in calibration and prediction. Therefore, we present a fast learning algorithm for artificial neural networks, which was introduced by Huang et al. in 2004. Our experimental results show that this approach is able to achieve good generalization performance with much less computational effort and is able to outperform the traditional gradient-based learning in artificial neural networks, which is a great advantage in analyzing high dimensional data. We demonstrate the capabilities of the new approach to genomic predictions in animal and plant breeding.
Keywords:
Genomic selection
Extreme learning machine
Complex traits
