Authors:
Matteo Hessel
1
;
Francesco Borgatelli
2
and
Fabio Ortalli
2
Affiliations:
1
Politecnico di Milano, Italy
;
2
TXT e-solutions, Italy
Keyword(s):
Model Tuning, Screening, Optimization, Machine-Learning, Adaptive Hill-Climbing, Sequential Masking.
Related
Ontology
Subjects/Areas/Topics:
Computer Simulation Techniques
;
Dynamical Systems Models and Methods
;
Formal Methods
;
Mathematical Simulation
;
Non-Linear Systems
;
Optimization Issues
;
Simulation and Modeling
;
Simulation Tools and Platforms
Abstract:
The aim of this paper is to describe a novel methodology for model-design and tuning in computer simulations, based on automatic parameter screening and optimization. Simulation requires three steps: mathematical modelling, numerical solution, and tuning of the model’s parameters. We address Tuning because, at the state-of-the-art, the development of life-critical simulations requires months to appropriately tune the model. Our methodology can be split in Screening (identification of the relevant parameters to simulate a system) and Optimization (search of optimal values for those parameters). All techniques are fully general, because they leverage ideas from Machine-Learning and Optimization Theory to achieve their goals without directly analysing the simulator’s mathematical model. Concerning screening, we show how Machine-Learning algorithms, based on Neural Networks and Logistic Regression, can be used for ranking the parameters according to their relevance. Concerning optimizati
on, we describe two algorithms: an adaptive hill-climbing procedure and a novel strategy, specific for model tuning, called sequential masking. Eventually, we show the performances achieved and the impact on the time and effort required for tuning a helicopter flight-simulator, proving that the proposed techniques can significantly speed-up the process.
(More)