Machine Learning
Begin with ML Foundations. For prediction, practise the shared Supervised Workflow, then choose a model deck for your task. For discovery or PCA, take Discovery foundations and the relevant preparation ideas. Helpful prior knowledge guides your choice; every lesson and workflow remains open.
Start with the foundations
Predict a quantity
After the supervised workflow, choose a model for a numerical outcome. Each model deck keeps its own question, mechanism and limits.
Simple linear regression
Read a fitted line, distinguish RMSE from R², then inspect residuals and extrapolation limits.
Multiple linear regression
Use several inputs, interpret encoded coefficients and examine correlated predictors.
Polynomial regression
Build curved features and compare flexibility using training validation.
Regression trees
Read leaf-average predictions, apply a full tree workflow and control complexity.
Regression reviews and checkpoint
Retrieve evidence across quantity models, examine explanation limits and assemble the regression checkpoint.
Predict a class
Begin with shared classification evidence, then choose a classifier. Quantity-model decks are a separate branch.
Classification evidence
Read confusion, precision, recall and macro F1 before interpreting labels or probabilities.
Logistic classification
Connect logistic scores and probabilities to a complete workflow and cost-sensitive thresholds.
Nearest-neighbour classification
Understand neighbour voting and validate k on a meaningful scale.
Support vector classification
Read margins, build a guided workflow and compare nonlinear boundaries.
One-R classification
Learn a one-feature rule with categorical or numeric inputs and inspect what it ignores.
Classification trees
Use class evidence in threshold rules and diagnose overfitting or minority-class failures.
Naive Bayes and input variants
Distinguish Gaussian measurements from Bernoulli flags; match the likelihood to the input and inspect independence assumptions.
LDA and QDA
Compare shared and class-specific covariance, then examine sample support and regularisation limits.
Classification reviews and checkpoint
Retrieve classification evidence across model families and complete the class-prediction checkpoint.
Neural networks
Learn the shared network ideas, then choose classification or quantity prediction. Neither task route sends you through the other.
Shared neural foundations
Learn outputs, loss and convergence, retrieve the shared ideas, then optionally explore capacity and regularisation.
Neural classification
Apply shared network ideas to class prediction and complete its independent checkpoint.
Neural quantity prediction
Apply shared network ideas to quantities in original target units and complete its independent checkpoint.
Discover structure
Start with a discovery question and meaningful measurements. Clustering and PCA use no prediction target or supervised final test.
Discovery foundations
Frame a question without a prediction target, then decide what distance and scale mean for these observations.
K-Means
Fit centroids, compare k, describe original-unit profiles and test the limits of K-Means geometry.
Hierarchical clustering
Inspect Ward merges, compare cuts and explain the scope of a sampled hierarchy.
Discovery interpretation and checkpoint
Interpret discovery without inventing true classes, retrieve hierarchy evidence and complete the discovery checkpoint.
Principal component analysis
Fit reusable axes, choose retained variation and interpret weights, scores and the limits of a two-dimensional view.
Compare and explain
Bring the model families together when you are ready to defend a choice and communicate its evidence.
Independent workflow challenges
Browse the original subject libraries
These links retain the complete subject lists, including shared reviews and checkpoints.
ML Foundations · Supervised Workflow · Regression · Classification · Neural Networks · Clustering and Discovery · PCA · Choose and Explain Models