Estimation and optimization of the performance of complex systems
This course is intended for general engineers and will provide:
- An understanding of the principles of stochastic algorithm methods (eg Monte Carlo, particulate, etc.)
- A choice of methods best suited to the problem of estimation or optimization being considered
- Techniques to put into practice for realistic applications
The course lasts 4 days (24 hours) and includes:
- 12 hours of theory-based learning with applications and examples
- 12 hours of PC-based application enabling learners to put into practice the methods taught for aerospace test cases
The usual estimation or optimization techniques are very quickly surpassed for complex and/or large-scale engineering problems such as the design of space launchers, the estimation of aircraft reliability or the optimization of the positioning of sensor networks.
Stochastic Monte-Carlo algorithms, inspired by games of chance, offer an efficient alternative for estimating and optimizing random parameters or calculating integrals efficiently. The development of these algorithms is currently fast-growing and many variants are available.
This course therefore offers an algorithmic and applicative overview of stochastic techniques and particle simulation methods. A day of training includes a morning course to present the different algorithms and an afternoon of practical PC-based work to implement and test these techniques on concrete problems from the aeronautics and space sectors.
Course level: Basic/Advanced
Engineering-level or equivalent training required.
Statistical knowledge is reviewed quickly at the beginning of the course.
Basic knowledge of Matlab and Python is required.
Senior Research Fellow at ONERA (Department of Information Processing and Systems).
(In connection with aeronautical and space applications)
- Summary of probability and statistics
- Classic Monte Carlo method
- Variance reduction methods
- Bootstrap technique
- Summary of Markov chains
- Markov Chain Monte Carlo Methods (MCMC)
- Multi-level simulation (subset simulation)
- Sequential Monte Carlo methods
- Particulate filtering (Kalman, SIS, SISR, etc.)
- Summary of optimization
- Simulated annealing, genetic algorithms, particle swarm optimization
- Probability law restart algorithms (CMA-ES, cross entropy)
- Stress management, multi-objective optimization
Scheduled in French:
TOULOUSE: 27 to 30 September 2021
For the English realization, please, consult us.
€2,200 excluding tax (20% VAT)