Figure 11.

Effect of regularization strength λ. Here λ = √λ_r is shown on the horizontal axis. (a) Model error X_d for all acceptable parameter sets for increasing λ, where light grey indicates an acceptable data fit and dark grey an unacceptable data fit. (b) Regularization error X_r for increasing λ, resulting in a so-called Pareto frontier. Any point on the frontier curve is Pareto efficient and it is up to the user to make trade-offs and select which λ is used. The dashed line indicates the λ that was selected (λ_r = 0.01). Adapted from Tiemann et al. [10].