JCMoptimizer is our advanced analysis and optimization toolkit. Machine learning technologies enable the efficient analysis and optimization of the properties of expensive blackbox functions such as FEM simulations of optical devices or real-world experimental setups.
Optimization: Bayesian optimization is a highly efficient optimization method that enables to develop high-performance devices in shorter computation times. Other supported optimization methods include downhill simplex optimization, particle swarm optimization, differential evolution, and the L-BFGS-B method.
Model calibration: Reconstructing system parameters like material properties and shape parameters from measured data is a complex numerical task. JCMsuite includes dedicated tools for the time efficient and precise reconstruction of parameter values and their measurement uncertainties.
Sensitivity analysis, active learning, etc.: A strong foundation on advanced machine learning models like Gaussian processes and neural network ensembles enables a wide range of applications ranging from active learning of fast surrogate models, over local or global sensitivity analyses to multi-objective optimization.
