Learning path
Next moduleModule 6: Trees and ensembles
Module 5
Instance-based and probabilistic models
Nearest neighbors, Naive Bayes, and support vector machines, three very different ideas of how to classify.
- 1. K-nearest neighborsPredict from the most similar stored examples, compute distances by hand, and see why feature scaling can flip a prediction.35 min
- 2. Naive BayesClassify with Bayes' rule by combining a class prior with per-feature likelihoods, compute a spam posterior by hand, and know when the independence assumption hurts.30 min
- 3. Support vector machinesFind the separating boundary with the widest margin, trade margin for errors with C, and bend the boundary with kernels.30 min
Three classic classifiers, each built on a different idea. KNN predicts from similar stored examples. Naive Bayes combines a prior with per-feature likelihoods using Bayes' rule. SVMs find the boundary with the widest margin and use kernels to bend it.
| # | Lesson | You will be able to |
|---|---|---|
| 1 | K-nearest neighbors | compute distances, vote, and show why scaling matters |
| 2 | Naive Bayes | compute a posterior by hand and explain the independence assumption |
| 3 | Support vector machines | explain margins, support vectors, the C trade-off, and kernels |
Interactive lab: K-nearest neighbors