My research interests focus on the development and implementation of methods that allow the probabilistic investigation of complex system dynamics. Basically, I
am working in the following four overlapping methodological domaines, often in the context of time-consuming determinisitic simulation codes and system observations or models consisting of a large
amount of data.
Machine Learning Algorithms
Combining machine learning algorithms to perform
- regression tasks to develop meta models for complex system dynamics, e.g. via Gauss Processes, Support Vector Machines
- adaptive sampling approaches to investigate complex system behavior in the vicinity of critical state transitions
Uncertainty & Sensitivity Analysis
Using advanced analysis approaches to investigate complex systems with focus on system safety and risk measures
- via Monte Carlo Dynamic Event Trees (MCDET) to exploratively study and probabilistically evaluate the impact of stochastic state transitions on the system dynamics
- via Uncertainty Analysis methods, i.e. estimation of Tolerance Limits via Wilks Approach, Bootstrapping or classical Monte Carlo
Simulations
- via Sensitivity Analysis methods to identify major contributors to the response uncertainty of a system, i.e. by squared multiple correlation coefficient or Sobol Indices
Bayesian Inference
Developping Bayesian inference approaches to
- detect and characterize multiple transition patterns in complex time series, such as paleo-climate observations via Linear Mixed Models
- localize cliff-edge effect regions via Gauss Processes
Complex Network Analysis
Using network analysis and representation approaches to visualize, investigate and compare
- complex dependency patterns, e.g. multiple hazard impacts on complex technical systems
- dynamic patterns of probabilistic risk models, e.g. to evaluate critical event sequences