Aller à : contenu haut bas recherche
 
 
EN     FR
Vous êtes ici:   UNIL > HEC Inst. > HEC App. > SYLLABUS
 
 

Advanced Data Analysis

  • Teacher(s):  
  • Course given in: English
  • ECTS Credits:
  • Schedule: Spring Semester 2017-2018, 4.0h. course (weekly average)
      WARNING :   this is an old version of the syllabus, old versions contain   OBSOLETE   data.
  •  sessions
  • site web du cours course website
  • Related programmes:

 

Objectives

The objective of the course is to gain practical familiarity with current computer aided data analysis and machine-learning approaches. Starting from classical regression introduced earlier, we will investigate recent statistical learning techniques, go beyond linear models and explore unsupervised approaches.

In the second part of the course we will turn to deep learning techniques and develop, evaluate and optimise neural network models using current implementation frameworks such as Keras and Tensorflow.

Contents

Statistical Learning

  • Resampling Methods, Cross Validation, Bootstrap
  • Selection and Regularisation, Ridge Regression, Lasso, Dimension Reduction
  • Beyond Linear Models: Polynomial Regression, Splines, Generalised Additive Models (GAM)
  • Tree-Based Methods: Regression & Classification Trees, Bagging, Random Forrests, Boosting
  • Support Vector Machines (SVMs), Support Vector Classifier, Multi-Class SVMs
  • Unsupervised Learning: Principle Component Analysis, K-Means, Hierarchical Clustering

Deep Learning

  • Basic Neural Networks, Activation Functions, Hyperparameters, Gradient-Based Learning, Back-Propagation
  • Regularisation for Deep Learning, Training Optimisation, Parallelisation and GPU Usage
  • Convolutional Networks (CNNs), Recurrent and Recursive Networks(RNN), Long Short-Term Memory Networks (LSTM)
  • Vectorisation, Selecting Hyper Parameters, Debugging Help & Techniques
  • Applications: Computer Vision, Speech Recognition, Natural Language Processing

Programming: students will learn to develop analysis programs in "R" using the popular RStudio environment to apply the concepts and tools to real data. During the practical part, we will put some emphasis on interactive, graphical data analysis and the use of R workbooks to create comprehensive and repeatable research reports.

References

  1. An Introduction to Statistical Learning (with Applications in R), James, Witten, Hastie, Tibshirani, Springer
  2. Deep Learning, Goodfellow, Bengio, Courville, MIT Press
  3. R for Data Science: Import, Tidy, Transform, Visiualize and Model Data, Wickham and Grolemund, O'Reilly

Pre-requisites

  1. Basic econometrics
  2. Datascience for finance

Evaluation

First attempt

Exam:
Written 2h00 hours
Documentation:
Not allowed
Calculator:
Not allowed
Evaluation:

Retake

Exam:
Written 2h00 hours
Documentation:
Not allowed
Calculator:
Not allowed
Evaluation:


[» go back]           [» courses list]
 
Search


Internef - CH-1015 Lausanne - Suisse  -   Tél. +41 21 692 33 00  -   Fax +41 21 692 33 05
Swiss University