MA2823: Introduction to Machine Learning

Course Overview

General Info: 2nd year course, CentraleSupélec, Fall 2018

Lecture hours: Friday, 13:45 - 17:00
Instructor: Fragkiskos Malliaros
Email: fragkiskos.me [at] gmail.com
Office hours: Right after class (or send me an email and we will find a good time to meet)

TAs: Dr. Maria Vakalopoulou (maria.vakalopoulou [at] centralesupelec.fr), Enzo Battistella (enzo.battistella [at] centralesupelec.fr), Mihir Sahasrabudhe (mihir.sahasrabudhe [at] centralesupelec.fr)


Piazza: piazza.com/centralesupelec/fall2018/ma2823/home


Machine learning is the scientific field that provides computers the ability to learn without being explicitly programmed (definition by Wikipedia). Machine learning lies at the heart of many real-world applications, including recommender systems, web search, computer vision, autonomous cars and automatic language translation.

The course will provide an overview of fundamental topics as well as important trends in machine learning, including algorithms for supervised and unsupervised learning, dimensionality reduction methods and their applications. A substantial lab section will involve group projects on a data science competition and will provide the students the ability to apply the course theory to real-world problems.





Schedule and Lectures

The topics of the lectures are subject to change (the following schedule outlines the topics that will be covered in the course). The slides for each lecture will be posted in piazza just before the start of the class. The due dates of the assignments/project are subject to change.


Week Date Topic Material Assignments/Project
1September 21IntroductionLecture 1
2September 28Dimensionality reductionLecture 2
3October 5Model selection and evaluationLecture 3Assignment 1 out
4October 12Linear and logistic regressionLecture 4 Project proposal due on October 14
5October 19Probabilistic classifiers and linear discriminant analysisLecture 5Assignment 2 out
Assignment 1 due on October 21
6November 9Non-parametric learning and nearest neighbor methodsLecture 6
7November 16Tree-based methods and ensemble learningLecture 7
8November 23Support Vector MachinesLecture 8
9November 30Introduction to reinforcement learning
Guest lecture by Dr. Nikolaos Tziortziotis
Lecture 9Assignment 2 due on December 2
10December 7Neural Networks
Guest lecture by Dr. Maria Vakalopoulou
Lecture 10
11December 14Unsupervised learning: clusteringLecture 11
12January 11ExamsProject final report due on January 6



[September 21] Lecture 1: Introduction

Introduction to machine learning, administrivia, course structure and overview of the topics that will be covered in the course. Bacic concepts in optimization.

Reading:
Additional:

[September 28] Lecture 2: Dimensionality reduction

Dimemensionality reduction techniques. Singular Value Decomposition (SVD). Principal Component Analysis (PCA). Multidimensional Scaling (MDS) and nonlinear dimensionality reduction.

Reading:
Additional:

[October 5] Lecture 3: Model selection and evaluation

Overfitting and generalization. Bias-variance tradeoff. Training, validation and test sets. Cross-validation. Evalution of supervised learning algorithms.

Reading:
Additional:
  • M. Kuhn and K. Johnson. An Introduction to Feature Selection. Applied Predictive Modeling, pages 487-519, 2013. [For the first part of the lecture on feature selection].
  • Model evaluation, model selection, and algorithm selection in machine learning: Part I (Basics), Part II (Bootstrapping and uncertainties), and Part III (Cross-validation and hyperparameter tuning). Interesting blog post by Sebastian Raschka, 2016.

[October 12] Lecture 4: Linear and logistic regression

Supervised learning models. Linear regression. Linear classification models. Logistic regression. Maximum likelihood estimation.

Reading: Additional:

[October 19] Lecture 5: Probabilistic classifiers and linear discriminant analysis

Bayes rule. Naive Bayes classifier. Maximum a posteriori estimation. Linear discriminant analysis (LDA).

Reading: Additional:

[November 9] Lecture 6: Non-parametric learning and nearest neighbor methods

Introduction to non-parametric learning methods. Distance and similarity metrics. Nearest neighbor algorithms.

Reading: Additional:

[November 16] Lecture 7: Tree-based methods and ensemble learning

Decision trees. Ensemble learning. Bagging and Boosting. The AdaBoost algorithm.

Reading: Additional:

[November 23] Lecture 8: Support Vector Machines

Maximum margin classifier. Linear SVMs. Primal and dual optimization problems. Non-linearly separable data and the kernel trick. Regularization and the non-separable case.

Reading:
Additional:

[November 30] Lecture 9: Introduction to reinforcement learning

Guest lecture by Dr. Nikolaos Tziortziotis.

Reading:

[December 5] Lecture 10: Neural Networks

Introduction to neural networks. The perceptron algorithm. Multilayer perceptron. Backpropagation. Applications.

Reading:

[December 14] Lecture 11: Unsupervised learning: clustering

Introduction to unsupervised learning methods. Data clustering. Hierarchical clustering. k-means clustering. Spectral clustering.

Reading:
Additional:


Course Structure and Objectives

Structure
Each section of the course is divided into 1h30' lecture and 1h30' lab. The labs will include hands-on assignments (using Python) and will provide the students the opportunity to deal with ML tasks in practice.


Learning objectives

The course aims to introduce students to the field of machine learning by:
  • Covering a wide range of topics, methodologies and related applications.
  • Giving the students the opportunity to obtain hands-on experience on dealing with.
We expect that by the end of the course, the students will be able to:
  • Identify problems that can be solved using machine learning methodologies.
  • Given a problem, identify and apply the most appropriate algorithm(s).
  • Implement some of those algorithms from scratch.
  • Evaluate and compare machine learning algorithms for a particular task.
  • Deal with real-world data challenges.



Prerequisites

There is no official prerequisite for this course. However, the students are expected to:

  • Have basic knowledge of probability theory and linear algebra.
  • Be familiar with at least one programming language (e.g., Python or any language of their preference).



Reading material

There is no single requiered textbook for the course. We will recommend specific chapters from the following books:



Evaluation

The evaluation of the course will be based on the following:

  1. Two assignments: the assignments will include theoretical questions as well hands-on practical questions that will familiarize the students with basic machine learning tasks.
  2. Project: The students are expected to form groups of 3-4 people, propose a topic for their project, and submit a final project report. Please, read the project section for more details.
  3. Final exam: Final exam in the material covered in the course.

The grading will be as follows:

Assignment 1 (individually): 10%
Assignment 2 (groups of 3-4 students): 20%
Project (groups of 3-4 students): 35%
Final exam: 35%


Academic integrity

All of your work must be your own. Don't copy another student's assignment, in part or in total, and submit it as your own work. Acknowledge and cite source material in your papers or assignments.



Project

Details about the project of the course have been posted on piazza.




Resources

Datasets


Software tools


Related conferences
Please find below a list of conferences related to the contents of the course (mostly in the field of machine learning and data mining. We provide the DBLP website of each venue where you can access the proceedings (papers, tutorials, etc).

Check out the website of each conference (e.g., KDD 2016 ) for more information.