Optimize or calibrate a model

The Optimization Studio separates reusable payoff definitions from optimization runs.

Define a payoff

  1. Open Model Optimization and select Payoffs.
  2. Add a payoff and choose its purpose.
  3. Add one or more model outputs.
  4. Set each output's evaluation window and weight.
  5. For behavior reproduction, select model references or import a dataset whose series names match the payoff elements.

Configure and run optimization

  1. Select Optimization Runs and add an analysis.
  2. Choose the payoff, search method, and Maximize or Minimize goal.
  3. Set the evaluation budget and number of saved best runs.
  4. Add parameters with baseline, minimum, and maximum values.
  5. Choose Run Optimization.

Available methods may include guided, evolutionary, and local searches. A larger evaluation budget can improve exploration but increases runtime. Start with narrow, defensible parameter ranges and a modest budget.

Interpret results

The lowest or highest score is only best for the payoff you defined. Check parameter plausibility, inspect the resulting trajectories, rerun validation, and test the result against data not used for calibration.

A completed behavior-reproduction optimization can provide R² and MAPE evidence to Validate Model. Results are also exported as CSV.

Yano Simulator Help — Version 0.1.0, Build 20260704