Lecture 1.5 — Three types of learning [Neural Networks for Machine Learning]

3-Second Blueprint (English)

1. Supervised learning tunes weights to minimize output error against labeled targets, handling real-valued regression or discrete classification.
2. Reinforcement learning selects action sequences under delayed feedback to maximize the time-discounted sum of expected future rewards.
3. Unsupervised learning extracts compact low-dimensional, binary, or sparse feature codes from raw inputs, with clustering as a one-hot limit.