ProteoAnalyst uses Monte Carlo random sampling to balance these groups during training. For instance, when you have unbalanced data (e.g., 20 diseased vs. 100 healthy), the algorithm creates many “mini-batches” of 20 vs. 20 for training and cross validation