From Statistical Inference to Learning Methods

Speaker : Freedath DJIBRIL MOUSSA (University of Abomey-Calavi,Benin)

This course revisits fundamental concepts of probability and statistics with a focus on methods and tools essential for statistical learning. Participants will strengthen their understanding of probabilistic modeling, stochastic dependencies, and inferential techniques in multivariate and high-dimensional contexts. Applications in diverse domains such as health, engineering, social sciences, and general data-driven decision-making will illustrate theoretical concepts. This course bridges solid statistical foundations with modern statistical learning methods, preparing participants for advanced courses on classification, regression, and machine learning.

This course introduces participants to key statistical and algorithmic methods for prediction and classification, bridging traditional statistical inference and modern learning techniques. Topics include supervised learning, model evaluation, and classification algorithms such as logistic regression, k-nearest neighbors, and decision trees. Emphasis is placed on practical applications in finance and other domains where predictive modeling is essential. Participants will learn to implement these methods in R or Python, interpret model outputs, and evaluate performance using cross-validation and error metrics. This course prepares participants for more advanced topics in statistical learning explored later in the school.

Speaker : Benoît LIQUET (Université de Pau et des Pays de l'Adour,France)

This course provides a robust introduction to statistical learning methods, bridging classical statistical techniques with modern machine learning approaches. Participants will explore supervised learning, unsupervised learning, and model evaluation methods, focusing on applications in finance, insurance, and other domains where predictive modeling is essential. Emphasis is placed on understanding algorithms from a statistical perspective, ensuring participants gain both theoretical and practical insights. Hands-on exercises using R and Python will help consolidate the concepts.

Censored and truncated data frequently arise in survival analysis, reliability studies, and financial risk modeling. Traditional statistical methods (e.g., Kaplan-Meier estimator, Cox proportional hazards model) provide strong tools, but recent advances in machine learning have opened new avenues for handling high-dimensional covariates, nonlinear effects, and complex censoring mechanisms. This course introduces modern machine learning approaches for censored data, combining theoretical insights with practical implementations. Topics include extensions of regression models to censored data, survival trees and forests, boosting methods, and neural network-based survival models. Applications will be drawn from credit risk, insurance, and biomedical studies.

Speaker : Gero JUNIKE (LMU Munich,Germany)

This course explores the integration of machine learning methods into financial mathematics, focusing on how advanced statistical learning techniques can be applied to problems in pricing, risk management, portfolio optimization, and financial forecasting. Participants will learn to bridge classical stochastic and mathematical finance with modern machine learning algorithms. The course will emphasize both theoretical underpinnings and practical applications through real financial data.

Speaker : Benoite DE SAPORTA (University of Montpellier,France)

This course introduces stochastic control and sequential decision-making. The focus is on modifying the natural trajectory of a stochastic process to optimize an objective function. The course covers Markov Decision Processes (MDPs), dynamic programming, solution algorithms, and extensions to partially observed problems or reinforcement learning for unknown models.

Speaker : El-Hadj DEME (University Gaston Berger,Senegal)

This course introduces methods for analyzing and modeling rare and extreme events using Extreme Value Theory (EVT) combined with machine learning techniques. Topics include block maxima, peaks-over-threshold methods, tail index estimation, and dependence modeling. Applications will be illustrated in finance, insurance, and environmental sciences. Hands-on sessions will cover the implementation of EVT estimators and machine learning algorithms (R/Python).

Implementation of random survival forests, CoxBoost, and DeepSurv using R (packages: survival, randomForestSRC, glmnet) and Python (lifelines, scikit-survival, pycox), hands-on examples: patient survival datasets, ecological survival studies, reliability datasets from engineering

Speaker : Gero JUNIKE (LMU Munich,Germany)

Implementation of ML models (regression, neural networks, hybrid models) on financial datasets (stock returns, option prices, etc.) using Python (TensorFlow/PyTorch, Scikit-learn).

Speaker : Benoite DE SAPORTA (University of Montpellier,France)

Simulation of stochastic processes and MDPs. Solving example decision problems using dynamic programming. Applying reinforcement learning algorithms to real or simulated datasets using in R/Python.

Speaker : El-Hadj DEME (University Gaston Berger,Senegal)

Extreme Value Theory estimation and Machine Learning models for extreme events in finance/insurance using R/Python.