The UCLouvain Doctoral Schools in Statistics, Biostatistics and Actuarial Sciences and in Computational Intelligence and Learning and the Statistical methodology and computing platform jointly organise a short course on PLS regression methods and extended tools with application to -omics data by Philippe Bastien (L'Oréal Research and Innovation, France)
March 14th and 15th 2016 - 9:00 am to 5:30 pm
Louvain-la-Neuve
Summary
PLS regression is a regularized robust alternative to multiple linear regression when dealing with multicollinearity. PLSR is very useful in particular in the analysis of wide data tables (p> n).
This course which aims to be pedagogic through many applications in R will provide basic elements useful to the understanding of the methods from a theoretical point of view.
PLS regression will be compared to other regularized alternatives such as principal component regression, ridge regression, Lasso and generalized Lasso. Extensions of PLS regression to discriminant analysis (PLS-DA Barker& Rayens), multiblocs, sparsity (variable selection), nonlinearity using kernels or splines, and generalized linear models will be fully described.
Course outline
- Regression and multicolinearity
- Regularization
- PLS Regression and Factorial methods
- PLS Regression and optimization
- SIMPLS
- NIPALS and missing data
- PLS1/PLS2 regression
- PLS DA
- OPLS
- Sparse PCA / Sparse PLS / Sparse PLS DA
- Other regularized methods: LASSO/ELASTICNET/GROUP LASSO
- Univariate soft thresholding & Cyclical coordinate descent algorithm
- PLS regression with high dimensional and low sample size : PLS Kernel
- Generalized PLS regression (PLSLogistique, PLSPoisson, PLSCox)
- Introduction to multiblocs PLS: Regularized Generalized Canonical Correlation Analysis
- Applications to omics data using R with packages: mixomics, glmnet, spls, rgcca, plsRglm,…
Prerequisite
- Theoretical and practical knowledge in projection methods like Principal Components Analysis
- Culture in multivariate statistical analysis, machine learning methods, o-mics methods
- Good knowledge in R programming
Price and registration
- Free for academics
- 300€ for non-academics (an invoice will be sent after registration)
- Number of participants limited.
Registration: http://www.uclouvain.be/en-627996
Information
Further information: edt-stat-actu@uclouvain.be
Room
C045 – LSBA - 20 voie du roman pays – 1348 Louvain-la-Neuve
Venue:
How to get to Louvain-la-Neuve ?
Map of Louvain-la-Neuve here 
If you want to stay in Louvain-la-Neuve overnight, you can book a room at
- Hotel IBIS. (Booking has to be done as soon as possible!
)
Boulevard de Lauzelle 61 - 1348 Louvain-la-Neuve
Tel: (+32)10/450751
@: h2200@accor.com - Hotel Leonardo (with public transport to LLN)
Rue de la Wastinne 45 B - 1301 Wavre
Tel: (+32)10/411363
@: info.wavre@leonardo-hotels.com - Hotel At Home (with public transportation to LLN)
Place Bosch 33 B - 1300 Wavre
Tel: (+32)10/228383
@: contact@at-homehotel.be - Hotels in Brussels (30km/45' by train)