← FixMatch Simplifying Semi-Supervised Learning with Consistency and Confidence Mean teachers are better role models Weight-averaged consistency targets improve semi-supervised deep learning results→