Fig. 5From: A biological network-based regularized artificial neural network model for robust phenotype prediction from gene expression dataFigure shows the impact of the choice of penalty on model performance. The bar plots indicate the average cross-validation balanced accuracy in multiple runs. In all experiments a regularization of the form r 1−r 2 has been applied where r 1 indicated the regularization applied to the weights in the first layer W (1) and r 2 indicates the regularization applied to the weights in the second layer W (2). Half L2: ℓ 2-Null), Half L1: Null- ℓ 1, Full L1: ℓ 1- ℓ 1, Full L2: ℓ 2- ℓ 2 and GRRANN: ℓ 2- ℓ 1 Back to article page