Core ML Concepts · intermediate · concept 42 of 311 · Updated Nov 4, 2026

📈Curriculum Learning

Training a model on easier examples first and harder ones later, like a school syllabus. Bengio and colleagues named the idea in 2009 and showed that a staged order speeds training and improves generalization. The hard part is deciding what counts as easy; self-paced and automatic curricula let the learner's own loss decide. OpenAI's robot hand learned to solve a Rubik's cube by widening its simulator's randomization only as it succeeded.