2026.2 — Overview

Welcome to the 2026.2 offering of Machine Learning.

Meetings


Class	Tue.	12h00 14h00
Class	Thu.	12h00 14h00
Office Hours	Mon.	14h45 16h15

Instructors

| Instructor | Humberto Sandmann |

Students

Syllabus

A — Foundations & History

#	Subject	Topics
1	Introduction & History	What is ML, a timeline from least squares (1805) to foundation models
2	The ML Landscape	Learning paradigms, the ML workflow, generalization, ethics

B — Working with Data

#	Subject	Topics
3	Exploratory Data Analysis	Summary statistics, distributions, correlation, visualization
4	Data Preprocessing	Scaling, normalization, encoding, missing values, outliers
5	Pipelines	scikit-learn `Pipeline`, `ColumnTransformer`, reproducibility

C — Unsupervised Learning & Text

#	Subject	Topics
6	Dimensionality Reduction	PCA, t-SNE, UMAP
7	Clustering	k-means, hierarchical, DBSCAN/HDBSCAN, silhouette
8	Text Representation	Bag-of-words, TF-IDF, n-grams, embeddings
9	Topic Modeling & BERTopic	LDA, BERTopic: embeddings + UMAP + HDBSCAN + c-TF-IDF

D — Regression & Model Evaluation

#	Subject	Topics
10	Linear Regression	Least squares, OLS derivation, assumptions, regression metrics
11	Gradient Descent & Regularization	Batch/stochastic GD, polynomial features, Ridge, Lasso
12	Validation & Data Leakage	Train/test split, cross-validation, leakage patterns
13	Model Selection	Bias–variance, regression to the mean, `GridSearchCV`

E — Classification

#	Subject	Topics
14	Classification & Metrics	Confusion matrix, accuracy, precision, recall, F1
15	k-Nearest Neighbors	Distance metrics, choosing k, curse of dimensionality
16	ROC-AUC & Imbalanced Data	ROC/PR curves, resampling, SMOTE, class weights
17	Logistic Regression	Sigmoid, cross-entropy, gradient descent, regularization
18	Naive Bayes	Bayes' theorem, conditional independence, spam filtering
19	Support Vector Machines	Margins, soft margin, kernel trick, implementation sketch

F — Trees & Ensembles

#	Subject	Topics
20	Decision Trees	Entropy, Gini, CART, pruning
21	Random Forest	Bootstrap, bagging, feature importance, out-of-bag error
22	Gradient Boosting	Boosting, GBM, XGBoost, LightGBM

G — Edge Approaches

#	Subject	Topics
23	Neural Networks	Perceptron to MLP, backpropagation, bridge to deep learning
24	Explainability	Permutation importance, SHAP, LIME
25	AutoML	Hyperparameter optimization, Optuna, successive halving
26	MLOps	Serving, monitoring, drift, reproducibility
27	The Frontier	Foundation models, transfer learning, LLMs, what's next

Class materials

Most lessons include a hands-on Colab notebook used in class (linked at the end of each lesson page, under Class materials). The complete collection — notebooks, slides, datasets, and papers — lives in the course Drive folder.

Assessment

Placeholder

Grading criteria, exam dates, and project descriptions will be published here at the start of the semester.