Objectifs

The objective of this advance data analytics course is to gain practical familiarity with current computer-aided data analysis and machine-learning approaches.

We will study methods from

a) supervised machine learning (classification and regression)

b) unsupervised machine learning

c*) reinforcement learning (if time permits)

A substantial part of the course will be devoted to learning on how to apply the current implementation frameworks TensorFlow, Keras, and scikit-learn and how to dispatch them on cloud-computing infrastructures by using Python.

In addition, we will have visitors from the financial and insurance industry holding guest lectures to demonstrate where these methods are applied in real-world settings.

Contenus

This course will be beginning with basic topics such as classification and linear regression and ending up with more recent topics from Deep Learning.

Some of the topics we will cover are:

• Linear Regression

• Classification

• Cross-Validation, MSE, BIAS

• Performance measures

• Regularization, Model Selection,

• Nonlinear regression

• Bayesian methods (e.g. Naive Bayes)

• Tree-based methods

• Support-vector machines

• Gaussian process regression and classification

• K-means, Mixture Models, and the EM Algorithm

• Dimension reduction (PCA, active subspace)

• Neural nets /Deep Learning

• Reinforcement learning

The hands-on examples will all be in Python.

Références

1. An Introduction to Statistical Learning (with Applications in R), James, Witten, Hastie, Tibshirani, Springer
2. Deep Learning, Goodfellow, Bengio, Courville, MIT Press
3. Pattern Recognition and Machine Learning, Bishop, Springer
4. Introduction to Machine Learning, Third Edition, Alpaydin, MIT Press

Pré-requis

1. Basic econometrics
2. Good programming know-how (in Python)
3. Basic calculus and probability

The course will consist of both lectures, software tutorials, and exercises. For the tutorials and exercises, you will need to bring a laptop computer to each class. If you do not have a laptop computer, you can still follow the class but please contact the professor to help you find a solution for effective participation in the practical hands-on exercises.

- For Master students from other universities that want to attend the course: Note that it is only dedicated to HEC master’s students and HEC mobility students (if you need credits; if you just want to follow the materials, there is no such restriction).

- For Ph.D. Students from other universities: Please write me an email to see whether you can attend the course for getting credit in your Ph.D. program.

Evaluation

1ère tentative

Examen:

Sans examen (cf. modalités)  

Evaluation:

There will be a graded home-take exam during the last week of the semester (48h time to solve it; 40% of the final grade) and a term paper of about 10 pages length that needs to be presented either in person, or via a pre-recorded video (depending on the Covid situation). The term paper and the presentation together account for the remaining 60% of the grade.

We will award the grades based on whether the home-take grades are generally on par with the class average and whether the final project and presentation demonstrate an understanding of the material.

Rattrapage

Examen:

Sans examen (cf. modalités)  

Evaluation:

Following HEC guidelines, we allow a second attempt to any or all of the partial grades with a result below 4.0 if and only if your overall grade is below 4.0. This means that, for example, if your project received a grade of 4.0 and your take-home exam a grade of 3.5, you have the choice to redo the take-home exam to try to move up the average grade.
If you received a grade below 4.0 for your project and choose to redo it, you will also have to present said project during the summer.

As during the semester, you will be granted 48h to complete and hand-in the retake of the take-home exam. Two weeks will be given to redo the project and present the results from a date agreed upon onward.

As with the regular evaluation, we will award the grades based on whether the home-take exam grade are generally on par with the class average and whether the final project demonstrates an understanding of the material.