Determining the impact of information loss across different image augmentations towards the performance of various convolutional neural networks on classifying mammograms