Parameters | Predictive Modeling | Compute near optimal low-rank randomized SVD for the additive and dominance matrices

Compute near optimal low-rank randomized SVD for the additive and dominance matrices
Specify this option to compute near optimal low-rank randomized singular value decomposition (SVD) of the additive and dominance matrices.
Randomized SVD is a powerful method to accurately compute low-rank SVD for large matrices, such as the additive and dominance matrices derived from data sets with many observations (individuals)1. Fitting a mixed model that contains large matrices A and D is computationally expensive if not prohibitive for very large matrices. Using randomized SVD is an alternative that speeds up computations without loosing accuracy in the predictions.
All parameters containing the same path as the output folder in the settings are also updated.
To Specify This Option:
*

1
Halko, N., Martinsson, P. G., and Tropp, J. A. (2011) Finding Structure with Randomness: Probabilistic Algorithms for Constructing Approximate Matrix Decompositions. SIAM Review, 53(2):217–288.