Statistics and data sciences

lepl1109  2020-2021  Louvain-la-Neuve

Statistics and data sciences
Due to the COVID-19 crisis, the information below is subject to change, in particular that concerning the teaching mode (presential, distance or in a comodal or hybrid format).
5 credits
30.0 h + 30.0 h
Q1
Teacher(s)
Hainaut Donatien; Jacques Laurent; SOMEBODY;
Language
English
Prerequisites
To follow this course the student must have a basic knowledge of probabilities such as taught in courses LEPL1108 or LBIR1212.

The prerequisite(s) for this Teaching Unit (Unité d’enseignement – UE) for the programmes/courses that offer this Teaching Unit are specified at the end of this sheet.
Main themes
This course presents the fundamental statistical concepts in an engineering context (exploratory analysis, inference, simulation) as well as basis method for analysing multivariate databases (like the linear regression, the principal component analysis and the classification).
Aims

At the end of this learning unit, the student is able to :

1
  • Explore datasets of small and big sizes with few or many dimensions
  • Infer features of a population from a sample using techniques of inference, estimation, confidence intervals and statistical tests.
  • To connect the deductive approach from the probability theory to the statistical inductive approach, and to identify the probabilistic models used in statistical inference. 
  • To translate the textual formulation of a problem of statistical inference into an accurate, statistical and mathematical formalism, while recognizing the adequate models and corresponding estimation methods.
  • To solve an applied problem by following a logical approach based on a correct use of models and statistical inference.
  • To use techniques of Monte-Carlo simulations, K-fold cross validation and bootstrapping in order to estimate models and validate results.
  • To analyse multivariate data with fundamental methods of linear regressions, of principal component analysis and of classification/clustering.
  • To use statistical tools to validate the conclusions from a model e.g. like the linear regression.
  • To make the link between the mathematical objectives of a method of data mining and its practical purposes.
 
Content
- Exploratory analysis and sampling
- Introduction to multivariate data analysis
- Parametric estimate (methods of moments and log-likelihood maximization) and properties of estimators (bias, variance, mean-squared error).
- Statistical inference (confidence intervals and significance tests): comparison of means of two or several normal populations, proportions, variance testing.
- Linear regression, including the analysis of coefficients and significance tests.
- Panorama of learning techniques, supervised and unsupervised learning methods
- Links between objectives of data analysis methods and their mathematical representation.
- Regression and classification methods (such as linear models and least square, k-nearest neighbors, logistic regression)
- Training, test error and generalization error, the Bias-Variance tradeoff, and elements of statistical decision theory
- Resampling techniques for model selection/evaluation (e.g., validation set, K-fold cross validation, bootstrap)
- Unsupervised learning: reduction of dimension (principal component analysis) and methods of clustering (K-means).
Teaching methods

Due to the COVID-19 crisis, the information in this section is particularly likely to change.

The course is composed of:
- 9 lectures on the topics listed in the course content;
- 7 practical sessions, both classical and numerical;
- 4 hackathons, representing 2 x 2 hours each, associated with small Python projects realized in group on subjects discovered both in the lectures and in the practical sessions.
Evaluation methods

Due to the COVID-19 crisis, the information in this section is particularly likely to change.

Written individual exam to evaluate the understanding of concepts and techniques   The hackathons represents 20% of the final mark. Lecturers keep the right to orally question students about their exam and hackathons.
Other information
To follow this course the student must have a basic knowledge of probabilities such as taught in courses LEPL1108 or LBIR1212. The schedule of course is subject to modifications due to sanitary conditions. Please check the Moodle website for more details.
Online resources
The totality of teaching material is available on the companion moodle website of the course. The schedule of course is subject to modification due to sanitary conditions, please consult the Moodle website of the course for additional information.
Faculty or entity
EPL
Force majeure
Teaching methods
- the lectures, the practical sessions as well as the hackathons are organized remotely
Evaluation methods
The evaluation is based on the continuous assessment and on a written exam, that respectively count for 20% and 80% of the final mark.  If sanitary conditions permit it, a closed book examination will be held on site, during the session. Failing this, a distance open book examination will be organized.


Programmes / formations proposant cette unité d'enseignement (UE)

Title of the programme
Sigle
Credits
Prerequisites
Aims
Bachelor in Computer Science

Bachelor in Engineering

Master [120] in Environmental Science and Management