Optimization (parameter estimation)

PEtab.jl is written to easily integrate with available optimization packages such as Optim.jl, Ipopt.jl and Fides.py. In the examples we show how to use these together with PEtab.jl.

Based on an extensive benchmark a good rule of thumb when choosing optimizer is:

• If you can provide a full Hessian the Interior-point Newton method in Optim.jl often outperforms the trust-region method in Fides.py.
• In case you cannot provide the full Hessian but the Gauss-Newton hessian approximation the Newton trust-region method in Fides.py often outperforms the interior-point method in Optim.jl.