Intra-Processing Methods for Debiasing Neural Networks