Insegnamento a.a. 2023-2024

• Basics of optimization
• Supervised learning: linear regression, logistic regression, kernel methods, SVM, decision trees, ensemble methods
• Neural Networks and backpropagation
• Bias-variance tradeoff, model selection
• Unsupervised learning: Principal Component Analysis, hierarchical clustering, k-means
• Basics of Reinforcement learning
• Implementing Machine Learning Pipelines with sklearn

• State and use the basic results of statistical inference for data analysis.
• Distinguish and characterize the following basic machine learning problems: regression, classification, clustering, dimensionality reduction.
• Describe the basic conceptual ideas, strengths, and limitations of the different learning algorithms for the settings above.
• Evaluate the main trade-offs such as overfitting, and computational cost vs accuracy.

• Implement end-to-end machine learning pipelines and rigorously evaluate their performance.
• Demonstrate the ability to cope with real-world data analysis and problem solving (managing, preprocessing and analyze real datasets).

• Face-to-face lectures
• Exercises (exercises, database, software etc.)
• Individual assignments
• Group assignments

• Face-to-face lectures focus on the theoretical, methodological, and computational aspects of the topics covered by the course.
• In hands-on exercise sessions students will work on their laptops to implement solutions to relevant case studies. Students will use Python and common libraries such as scikit-learn.
• Individual/group assignments: each student is required to solve a machine learning problem and provide a written report and the related code.

The exam consists of a theory part and a hands-on project.

There are two options to pass the exam:

1. Option 1: final exam for the whole 31 points. It will consist of questions (closed & open answer) and exercises to be answered on paper. The final exam is closed-book.
2. Option 2: final exam (same rules as above) + ML project

• The theory part consists in exercises and questions to be answered on paper. This contributes to 50% of the final grade.
• The project consists in programming code to solve a real ML problem, to be developed individually or in small groups and described through a written report, which is evaluated by the instructors. This contributes to 50% of the final grade.
• In order to pass the exam with Option 2, students must achieve a passing grade in both the theory and the project part (that is, at least 18 points out of 31 in each part). If the student is sufficient in both parts, then the grade is computed as 0.5 X + 0.5 Y, where X and Y are the score obtained in the final written exam and in the project, respectively
• The written exam is the same regardless of whether or not you participate in the challenge
• There is going to be one ML project for each edition of the course. If you participate in the project and obtain a passing grade, you can keep such grade and cumulate it with a final exam done in later sessions (it can be used up to the September session of the same year in which you did the project)

The written exam is used to asses the "knowledge and understanding" learning objectives. The project is used to asses the "applying knowledge and understanding" learning objectives.

The course mostly relies on the following two books (both available online for free): 

• Lindholm, A., Wahlström, N., Lindsten, F., & Schön, T. B. (2022). Machine learning: a first course for engineers and scientists. 
• James, G., Witten, D., Hastie, T., Tibshirani, R., & Taylor, J. (2023). An introduction to statistical learning: With applications in Python.

Other useful references:

• C.M. BISHOP (2006). Pattern Recognition and Machine Learning.
• T. HASTIE, R. TIBSHIRANI, J. FRIEDMAN (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction.

Slides and/or notes summarizing the contents presented in class are provided.