1.2 Supervised, Unsupervised, and Reinforcement Learning

🎯 Key AWS AI Exam Takeaways

• Supervised Learning uses labeled ground-truth pairs for Regression (continuous) or Classification (discrete).
• Unsupervised Learning discovers latent clusters and representations without ground-truth labels.
• Reinforcement Learning trains an autonomous agent to maximize cumulative rewards through environmental interaction.

Supervised learning requires training data consisting of both input features and corresponding ground-truth target labels. When the target variable is continuous (e.g., predicting dollar sales or temperatures), regression algorithms are used. When the target consists of discrete categories (e.g., fraud vs. legitimate), classification algorithms are applied.

Unsupervised learning operates on datasets devoid of pre-existing target labels. Algorithms such as K-Means clustering and Principal Component Analysis (PCA) discover hidden groupings, natural clusters, and dimensionality reductions based strictly on geometric distance and feature distributions.

Reinforcement learning models an agent taking sequential actions within an environment to transition between states. By receiving scalar rewards for favorable actions and penalties for undesirable states, the agent learns an optimal behavioral policy over time through trial and error.

⚠️ Common Pearson VUE Exam Traps

• Do not confuse clustering (unsupervised) with classification (supervised with known classes).
• Remember that continuous predictions (house prices) are regression, while discrete predictions (churn yes/no) are classification.