The knowledge map

AExplorer

• AAccuracy: how good is the model?
• ATraining a machine with examples

BInvestigator

• BClassification: how a model sorts information
• BNearest neighbour: classification based on similarity
• BRule-based systems and machine learning
• BWhen the model is wrong — and who it hits
• BYes/no questions that sort: decision trees

CBuilder

• CExperiment: biased training data and its consequences
• CGrouping without an answer key: clustering by hand
• CLinear regression: fitting a line to data
• CLinear regression: fitting a straight line to data
• COverfitting: the difference between learning the rule and memorising
• CProject: train an image classifier in the browser

DAI developer

• DClustering: k-means and hierarchical
• DCross-validation
• DDecision trees
• Dk-nearest neighbours (kNN)
• DLinear regression with several features
• DLogistic regression and decision trees
• DNaive Bayes
• DOverfitting and generalisation
• DPrecision, recall, F1 and ROC
• DRegression metrics: MAE, RMSE, R²
• Dscikit-learn — the workflow
• DTraining, validation and test

EUniversity

• EAnomaly detection
• EDimensionality reduction: PCA, t-SNE, UMAP
• EEnsembles: random forest and boosting
• EHyperparameter search
• ERecommender systems
• ESupport vector machines (SVM)
• EThe bias–variance trade-off
• ETime series