EE-612 / 4 crédits

Enseignant(s): Anjos André, Canévet Olivier, Marcel Sébastien

Langue: Anglais

Remark: Next time: Spring 2025

Every 2 years

## Summary

This course provides in-depth understanding of the most fundamental algorithms in statistical pattern recognition or machine learning (including Deep Learning) as well as concrete tools (as Python source code) to PhD students for their work.

## Content

This course will cover the broad regression, classification and probability distribution modeling methods and more particularly: Linear regression, Logistic regression, k-NN, Decision Trees, Boosting, Dimensionality reduction (PCA, LDA, t-SNE), k-Means, GMMs, MLPs, CNNs, SVMs.

A - Introduction

• Data representation,
• Pattern Recognition and Machine Learning,
• Lab preparation (JupyterHub, Python and pyTorch).

B - Regression and Classification

• Linear Regression,
• Logistic Regression and Regularization, Overfitting and Capacity,
• k-NN, Decision Trees,
• Artificial Neural Networks: Multi-Layer Perceptron (MLP) and Back-Propagation
• Deep Learning : Convolutional Neural Networks (CNN) and Optimization
• Support Vector Machines

C - Dimensionality reduction and Clustering

• Principal Component Analysis (PCA),
• Linear Discriminant Analysis (LDA),
• k-Means, Single Linkage,
• t-SNE.

D - Probability distribution modelling

• Gaussian Mixture Models (GMM) and the Expectation-Maximization (EM).

## Keywords

Pattern Recognition, Machine Learning, Linear models, PCA, LDA, MLP, SVM, GMM, HMM.

## Recommended courses

Linear algebra, Probabilities and Statistics, Signal Processing, Python (for the Labs).

## Assessment methods

Laboratory and oral exam.

## Dans les plans d'études

• Nombre de places: 24
• Forme de l'examen: Multiple (session libre)
• Matière examinée: Fundamentals in statistical pattern recognition
• Cours: 36 Heure(s)
• TP: 20 Heure(s)
• Type: optionnel
• Nombre de places: 24
• Forme de l'examen: Multiple (session libre)
• Matière examinée: Fundamentals of superresolution optical microscopy and Scanning Probe Microscopy
• Cours: 36 Heure(s)
• TP: 20 Heure(s)
• Type: optionnel

## Cours connexes

Résultats de graphsearch.epfl.ch.