Evaluation of Predictive Ability of Bayesian Regularized Neural Network Using Cholesky Factorization of Genetic Relationship Matrices for Additive and Non-additive Genetic Effects

This study aimed to explore the effects of additive and non-additive genetic on prediction complex traits using Bayesian regularized artificial neural network (BRANN). The data sets were simulated for two hypothetical pedigrees with five different fractions total variance accounted by additive, x effects. A feed forward (ANN) regularization (BR) was used assess performance nonlinear ANNs compare their predictive ability those from linear models under architectures phenotypic traits. Effective number parameters sum squares error (SSE) in test evaluate ANNs. Distribution weights correlation between observed predicted values set ability. There clear significant improvements terms (equivalent ridge regression) when proportion ( ) increased. On other hand, outperformed across architectures. larger more variable than network, presented leptokurtic distributions, indicating strong shrinkage towards 0. In conclusion, our results showed that: a) inclusion did not improve compared purely models, b) BRANN activation function substantially scenarios considered.