Modern Machine Learning

Create a free account to open this chapter.

• What machine learning is, vs. AI and deep learning ~8 min
• The supervised learning spectrum ~9 min
• Features, samples, and data splits ~8 min
• Generalization: the core goal ~8 min
• Models, training, inference, and prediction ~7 min
• The complete ML workflow ~10 min
• Iteration and ML experimentation ~7 min

Create a free account to open this chapter.

• Choosing a tool: library philosophy ~7 min
• NumPy vectorization and broadcasting ~8 min
• pandas: grouping, missing data, and joins ~9 min
• The scikit-learn API: fit, predict, transform ~8 min
• Where the rest of the ecosystem fits ~8 min

Create a free account to open this chapter.

• Visualizing one variable: histograms and box plots ~8 min
• Visualizing relationships: scatter plots and correlation ~8 min
• Distributions and skew ~7 min
• Class imbalance ~7 min
• Multicollinearity ~7 min
• Data leakage: a closer look ~8 min

Create a free account to open this chapter.

• Scaling and standardization ~8 min
• Encoding categorical variables ~8 min
• Missing-value imputation ~8 min
• Transforming skewed features ~7 min
• Feature creation ~8 min
• Feature selection ~7 min
• Pipeline and ColumnTransformer ~9 min

Create a free account to open this chapter.

• The validation-set approach and its weaknesses ~7 min
• K-fold cross-validation ~8 min
• Stratified k-fold ~6 min
• The bootstrap ~7 min
• Nested cross-validation ~8 min
• Reproducibility and random seeds ~6 min

Create a free account to open this chapter.

• Simple linear regression and least squares ~9 min
• Multiple linear regression ~9 min
• Evaluating a regression: R-squared, MAE, MSE, RMSE ~8 min
• Overfitting in linear models ~9 min
• Ridge regression ~9 min
• Lasso regression ~9 min
• Elastic net and the bias-variance tradeoff in practice ~8 min

Create a free account to open this chapter.

• Binary logistic regression ~9 min
• Multiclass logistic regression ~7 min
• Generative classifiers: LDA and QDA ~9 min
• Naive Bayes, revisited for classification ~8 min
• K-nearest neighbors ~8 min
• Distance metrics and the curse of dimensionality ~8 min
• Comparing classification methods ~8 min

Create a free account to open this chapter.

• Regression trees and recursive binary splitting ~9 min
• Cost-complexity pruning ~10 min
• Classification trees: Gini, entropy, and classification error ~9 min
• Trees vs. linear models ~7 min

Create a free account to open this chapter.

• Bagging and out-of-bag error ~8 min
• Random forests ~8 min
• Boosting: AdaBoost and gradient boosting ~9 min
• XGBoost, LightGBM, and CatBoost ~8 min
• Voting, stacking, and blending ~8 min

Create a free account to open this chapter.

• The maximal margin classifier ~9 min
• The soft margin and cost parameter ~9 min
• The kernel trick and support vector machines ~10 min
• Gaussian processes ~8 min

Create a free account to open this chapter.

• Precision, recall, and F1 ~9 min
• ROC curves, AUC, and threshold tuning ~9 min
• Grid search and random search ~8 min
• Bayesian optimization with Optuna ~8 min
• Learning curves ~8 min

Create a free account to open this chapter.

• K-means clustering ~8 min
• DBSCAN ~8 min
• Hierarchical clustering ~8 min
• Gaussian mixture models and the EM algorithm ~10 min
• PCA: two views of the same idea ~9 min
• Running PCA in practice ~8 min
• t-SNE and UMAP ~8 min

Create a free account to open this chapter.

• Statistical and model-based anomaly detection ~9 min
• Oversampling, SMOTE, and undersampling ~8 min
• Class weighting and threshold optimization ~8 min
• An imbalance sweep in practice ~8 min

Create a free account to open this chapter.

• Feature importance and permutation importance ~8 min
• Partial dependence and ICE plots ~8 min
• SHAP: explaining individual predictions ~9 min
• Probability calibration ~8 min
• Does 90% confidence mean 90% correct? ~8 min

Create a free account to open this chapter.

• Time series features: lags and rolling statistics ~8 min
• Data leakage in time series ~8 min
• Text as data: tokenization, bag-of-words, and TF-IDF ~8 min
• Text classification: Naive Bayes and logistic regression ~7 min
• Modern embeddings and semantic search ~8 min

Create a free account to open this chapter.

• From logistic regression to a neural network ~8 min
• Activation functions and forward propagation ~8 min
• Backpropagation: the chain rule you already know ~8 min
• What deep learning teaches next ~6 min

Create a free account to open this chapter.

• Why models fail: a troubleshooting order ~9 min
• Distribution shift, spurious correlations, and over-tuning ~8 min
• Bias and fairness in practice ~8 min
• Privacy, explainability, and dataset provenance ~8 min
• Model cards and responsible documentation ~6 min
• Saving a model: joblib and pickle caveats ~7 min
• ONNX and model serialization ~6 min
• Experiment tracking and the model registry ~7 min
• The MLOps loop and what comes next ~7 min

Create a free account to open this chapter.

• Defining the problem ~8 min
• Exploring and preparing the data ~8 min
• A baseline model, and why it's not optional ~6 min
• Trying multiple candidate models ~7 min
• Evaluating and comparing candidates honestly ~7 min
• Tuning the chosen model ~7 min
• Interpreting and saving the final pipeline ~8 min
• Real-world project archetypes ~7 min
• Which algorithm should I use? ~10 min