Common UDA technical approaches include adversarial training (e.g., the DANN method), Maximum Mean Discrepancy (MMD), and self-training/pseudo-labeling