Question

A medical imaging startup has 500,000 X-ray images, but only 12,000 have been labeled by radiologists due to the high cost of expert annotation. The team wants to build a diagnostic classification model that leverages both the small labeled set and the large unlabeled set to improve accuracy. Which machine learning approach best fits this situation?

Accepted Answer

Semi-supervised learning

Answer

Fully supervised learning using only the 12,000 labeled images

Answer

Unsupervised learning using clustering on all 500,000 images

Answer

Reinforcement learning with a reward signal per diagnosis