Multi-Goal Optimization

In view of the above-mentioned impossibility of simultaneously optimizing all exposure quality metrics (M1:strength, M2:selectivity, and M3:collateral; see Quantities-of-Interest), a Pareto multi-goal optimization approach is chosen, wherein an entire Pareto front of 'optimal' configurations is identified, from all possible electrode configurations and current ratios between the two channels. Here 'optimal' has the meaning that no configuration is known that improves on quality metrics, without degrading performance with regard to one or multiple other metrics.

Surrogate-Modeling-Based Optimizer (SuMo)

Starting with TIP V4.0, the optimization process uses a surrogate-modeling-based optimizer (SuMo) which combines Gaussian process modeling with a multi-objective genetic algorithm. This approach systematically explores the solution space to provide a diverse set of Pareto-optimal solutions, enabling users to efficiently navigate trade-offs like selectivity versus intensity.

The SuMo optimizer works by:

1. Building surrogate models of the objective functions using Gaussian process regression
2. Using these models to efficiently guide the search toward promising regions of the parameter space
3. Employing a multi-objective genetic algorithm to find diverse Pareto-optimal solutions
4. Continuously refining the surrogate models as more evaluations are performed
5. Running multiple independent seeds in parallel and merging their Pareto fronts via non-dominated sorting, yielding denser and more reliable front coverage
6. Monitoring the hypervolume indicator after every iteration and stopping each seed automatically once convergence is reached

This approach delivers high-quality, dense Pareto fronts with user-controllable precision (Low / Medium / High), and stops each optimization seed as soon as meaningful improvement ceases — avoiding unnecessary computation while ensuring thorough coverage of the trade-off surface.

From this front of solutions, a specific solution can either be chosen by focusing on a single metric, or by assigning weights to the three metrics. Both is possible in the second step of the planning workflow (Optimal Configuration Identification). Furthermore, the current ratio can be varied in this step, supported by graphs that show the current-ratio-dependence of M1–M3. The user is also free to interactively vary other parameters of the exposure configuration (e.g., electrode selection).