In this workshop we will focus on Bayesian inference as well as optimizationand sampling with application to the Lotka-Volterra predator-prey system ofnon-linear ordinary differential equations (ODEs). The first half of theworkshop will focus on the theoretical foundations for the problem of interestand the second half will apply them using the Korali software from a sequentialand parallel perspective through the Python programming language.Korali is a high-performance framework for Bayesian Uncertainty Quantification(UQ), optimization, and reinforcement learning. Korali's multi-languageinterface allows the execution of any type of computational model, eithersequential or distributed (MPI) using the C++ or Python programming languages.Korali provides a simple interface that allows users to easily describestatistical / deep learning problems and choose the algorithms to solve them.
Session 1 (THEORY):
What is Bayesian inference?
Optimization and sampling
Session 2 (THEORY/KORALI):
Bayesian inference for computational models
Inferring parameters for systems of ODEs
Application example: Lotka-Volterra (fit parameter with uncertainty)
Introduction to Korali and user interface via python
Session 3 (KORALI):
Practical examples in Korali
Running Korali in parallel