AI & ML
Foundations of Learning
Part of Foundations, level 0. About 8 hours of study across the topics below.
Orientation
- What Is a Model?Models as adjustable machines: parameters turn inputs into predictions, unlike fixed programs.1–2 h
- Learning from DataTraining fits parameters to data; inference uses fitted parameters on new inputs.1–2 h
- Tasks: Regression, Classification, and BeyondML tasks classified by output type: numbers, labels, sequences, and structures.1–2 h
Learning Framing
- Features and TargetsSamples as rows: feature vectors in, target out; the (x, y) contract behind every lesson.1–2 h
- Datasets and SplitsTrain/validation/test separation: fit here, judge there, and never peek.1–2 h
- Generalization PreviewFitting training data is not learning: the train/test gap, stated as intuition only.1–2 h
- Framing an ML Problem: Objectives, Baselines, and ProxiesBefore any model: what decision is served, which observable proxy is optimized, and what simple baseline must be beaten.1–2 h
- Learning Paradigms: Where the Supervision Signal Comes FromSupervised, unsupervised, self-supervised, and reinforcement learning distinguished by the source of the training signal.1–2 h