The definitive 14-module curriculum: from formal learning paradigms, statistical foundations, and data preprocessing to linear/logistic regression, regularization, decision trees & ensembles, SVMs, clustering, dimensionality reduction, evaluation metrics, and production scikit-learn pipelines.
Complete all 14 milestones to unlock and download your verified certificate (14 remaining).
Formal definition of machine learning, mathematical formulation, learning paradigms, and first hands-on models.
Generalization error, empirical risk minimization, underfitting, overfitting, and the bias-variance tradeoff.
Leak-proof data prep: fit-on-train rule, missing data imputation, robust scaling, and categorical encodings.
Simple/multiple linear regression, Ordinary Least Squares (OLS) derivation, gradient descent, and residual diagnostics.
Binary/multiclass logistic regression, log-odds, sigmoid activations, cross-entropy loss, and decision thresholds.
Controlling complexity and multicollinearity: L2 Ridge, L1 Lasso feature selection, and Elastic Net combinations.
Splitting criteria (Gini, Entropy), tree pruning, Random Forests, AdaBoost, and Gradient Boosted Trees (XGBoost).
Maximum margin hyperplanes, soft-margin optimization, dual Lagrangian, and non-linear kernel tricks (RBF).
Instance-based learning, curse of dimensionality, Bayes theorem, and probabilistic classifiers.
Unsupervised pattern discovery: K-Means optimization, Hierarchical clustering, DBSCAN, and Silhouette validation.
High-dimensional data compression: PCA via covariance eigendecomposition/SVD, and non-linear manifold t-SNE.
Confusion matrices, Precision-Recall tradeoff, ROC-AUC, F1-scores, and regression performance metrics.
K-Fold cross-validation, GridSearchCV, RandomizedSearchCV, Bayesian optimization, and leak-proof nested CV.
Custom transformers & estimators, ColumnTransformer pipelines, model serialization with Joblib, and deployment.