The Chair “Stress Testing” is a specific research program between the Ecole Polytechnique, BNP Paribas, Fondation de l'Ecole Polytechnique, and is hosted at l'X by the Center of Applied Mathematics. This research project is part of an in-depth reflection on the increasingly sophisticated issues surrounding stress tests (under the impulse of the upcoming European Banking regulation). Simulation of extreme adverse scenarios is an important topic to better understand which critical configurations can lead to financial and systemic crises. These scenarios may depend on complex phenomena, for which we partially lack information, making the modeling incomplete and uncertain. Last, the data are multivariate and reflect the dependency between driving variables. From the above observations, different lines of research are considered:
  1. Generation of stress test and meta-modeling scenarios using machine learning
  2. Quantification of uncertainties in risk metrics
  3. Modeling and estimation of multidimensional dependencies
Keywords: Bayesian Networks, copulas, dependent data, Deep Learning, Gaussian processes, machine learning, Markov Chain Monte Carlo, meta-modeling, multivariate statistics, rare event simulation, risk metrics, splitting methods, stochastic algorithms, stochastic and Bayesian optimization, uncertainty propagation

Ecole Polytechnique BNP Paribas Fondation de l'X