Introduction

This course introduces students to the core principles and methods of pattern recognition, with a particular emphasis on statistical learning techniques for classification (30 hours). It provides a solid theoretical foundation complemented by practical implementation using Python and the Scikit-learn library. Students will learn to evaluate classification models and apply them to real-world data.

  • Luis Baumela, 15h
  • Roberto Valle, 15h

Unit 1: Introduction to Pattern Recognition

This unit sets the foundations of the course:

  • Definition and scope of pattern recognition.
  • Motivation and real-world applications.
  • Euclidean distance classifier as a baseline approach.

Unit 2: Statistical Foundations of Classification

Focuses on the probabilistic perspective for classification:

  • Bayesian decision theory.
  • Parametric generative classifiers (e.g., Gaussian models).
  • Non-parametric classifiers (e.g., k-NN, Parzen windows).

Unit 3: Classifier Evaluation

Covers key methods to assess model performance:

  • Confusion matrix, accuracy, precision, recall.
  • Cross-validation techniques.
  • ROC curves and AUC.

Unit 4: Dimensionality Reduction

Techniques to manage high-dimensional data and improve model efficiency:

  • Feature selection vs. feature transformation.
  • Principal Component Analysis (PCA).
  • Linear Discriminant Analysis (LDA).

Unit 5: Discriminative Classifiers

Introduction to widely used discriminative models:

  • Logistic regression.
  • Support Vector Machines (SVM).
  • Decision boundaries and margin-based learning.

Unit 6: Unsupervised Classification

Explores clustering methods for structure discovery in unlabeled data:

  • K-means clustering.
  • Hierarchical clustering.
  • Evaluation of clustering quality.