Machine Learning Algorithms Ladder¶
Who This Is For¶
Use this ladder when you intentionally want one textbook-breadth spillover from algorithms into ML, while keeping the route algorithmic and narrow.
Warm-Up¶
- state the classifier as
sign(w · x + b) - explain what linearly separable means
- say exactly when the perceptron updates
Core¶
- deterministic perceptron
- mistake-driven online updates
- convergence only under separability
Repo Anchors¶
Stretch¶
- compare perceptron with logistic-regression style thinking without pretending they are the same route
- read one official course note on margins and explain what the repo lane still intentionally does not cover
Compare Points¶
- first algorithmic ML benchmark -> Perceptron Classification Benchmark
- online update worldview without statistical learning -> Online Algorithms
- probabilistic design is the actual topic instead -> Randomized Algorithms
This ladder is intentionally tiny.
The point is not to turn the repo into an ML curriculum. The point is to prevent this source-backed breadth topic from staying forgotten or fuzzy.
Exit Criteria¶
You are ready to move on when you can:
- implement the perceptron with bias correctly
- explain why separability matters for convergence
- say clearly what this lane does not cover