Machine Learning

Teaching objectives:

This subject is a deepening of the foundations of Machine Learning: Types of learning, data preprocessing, pattern extraction, supervised classification, cross-validation and model optimization. It allows students to:
● Learn to understand the data.
● Know and be able to apply the theoretical and practical fundamentals of Machine Learning.
● Know and be able to apply techniques for extracting knowledge from any given data.
Recommended prior knowledge: Mathematics: Linear algebra, probability, statistics, etc.

Content

  1. Introduction and Fundamentals
    ● Overview of Machine Learning.
    ● Types of ML and the lifecycle of a project.
    ● Applications and differences from Data Mining.

  2. Data Preprocessing
    ● Handling missing values and outliers.
    ● Normalization, discretization, encoding.
    ● The problem of imbalanced datasets.

  3. Regression
    ● Linear regression (simple and multiple).
    ● Regularization (Ridge, Lasso).

  4. Classification
    ● Simple perceptron (introduction to networks).
    ● Decision trees and forests.
    ● K-nearest neighbors.
    ● Naïve Bayes.
    ● SVM (linear separators).

  5. Clustering
    ● Similarity measures.
    ● K-means, HAC.
    ● DBSCAN.

  6. Model Evaluation and Validation
    ● Metrics for regression (MAE, RMSE, R²).
    ● Metrics for clustering (silhouette, Davies-Bouldin).
    ● Advanced validation techniques (stratified CV, nested CV).
    ● Cross-validation (K-fold, leave-one-out).
    ● Overfitting vs underfitting (learning curves).
    ● Hyperparameter optimization.
    ● Interpretation of results (accuracy, precision, recall, F1)

  7. Feature Engineering and Anomaly Detection
    ● Feature creation and selection.
    ● Dimensionality reduction (PCA).
    ● Detection methods (Isolation Forest, LOF).

  8. Frequent Patterns and Association Rules
    ● Basic concepts.
    ● Methods for finding frequent patterns.
    ● Transition to association rules.
    ● Sequential frequent patterns.

  9. Case Studies and AutoML
    ● Project 1: Advanced customer segmentation.
    ● Project 2: Fraud detection.
    ● Introduction to AutoML (TPOT, Auto-sklearn).*
    Evaluation method:
    ● In-person semester exam (60%).
    ● Continuous assessment (CC) (40%): Continuous tests, personal work.
    Bibliographic references:

1. Murphy, K.M., “Machine Learning”, MIT Press, 2012.
2. Mohri, M., Rostamizadeh, A., and Talwalkar, A., “Foundations of Machine Learning”, MIT
Press, 2012.
3. Goodfellow, I., Bengio, Y., and Courville, A., “Deep Learning”, MIT Press, 2016.
4. Borwein, J. M., and Lewis, A. S., “Convex Analysis and Nonlinear Optimization: Theory and
Examples”, Springer, 2006.
5. Practical MLOps, Noah Gift & Alfredo Deza, O'REILLY Media, 2021.
6. Introducing MLOps, Mark Treveil& the Dataiku Team, O'REILLY Media, 2020.