# Parameter optimization

To start parameter optimization, click on Tools->Optimization, or use the keyboard shortcut Ctrl+P.

The parameter optimization interface displays a history of the best set of parameters and their fitness, along with images of the optimization targets (species concentrations or concentration rate of changes) compared with the resulting values from the current best set of parameters.

• Targets and results (left side)
• If you have multiple targets, you can choose which one to display from the dropdown list

• The upper image shows the target concentration (or concentration rate of change) values

• The lower image shows the actual values from the simulation using the current best set of parameters

• Fitness and parameters (right side)
• The upper plot shows the history of fitness values during the optimization (lower is better)

• The lower plot shows the history of the best set of parameters for at each optimization iteration

• Setup
• Click the Setup button to change the optimization settings

• See the Optimization Setup section for more details

• Start
• Click the Start button to start optimizing

• You can stop and continue optimization

When you are happy with your parameters you can press Stop and then OK. You will then be asked if you want to apply the best parameters found during optimization to your model. The optimization settings (but not any existing optimization history or results) are also saved in the model.

## Optimization Setup

• Algorithm
• Choose which optimization algorithm to use

• See the Pagmo algorithms documentation for more details about the available algorithms

• The number of populations to evolve in parallel

• Typically this would be equal to the number of available CPU cores

• Population
• The population size to evolve on each thread

• Typically the more parameters are being optimized the larger this should be

• Targets
• A list of target concentrations (or rate of change of concentrations) to optimize for

• See the Optimization Target section for more details

• Parameters

## Optimization Parameter

• Parameter
• This can be a model parameter or a reaction parameter

• Lower bound
• The minimum allowed value this parameter can take

• Upper bound
• The maximum allowed value this parameter can take

## Optimization Target

• Species
• The species to target

• Target type
• Either the concentration, or the concentration rate of change

• Simulation time
• The timepoint in the simulation when this target should apply

• Target values
• The desired spatial distribution of values

• If not specifying this defaults to zero everywhere in the compartment

• See the Optimization Target Image Import section for more details

• Difference type
• How to compare the target $$t$$ with the results $$r$$

• Absolute: $$|r - t|$$

• Relative: $$|r - t|/|t + \epsilon|$$

• Weight
• The relative importance of this target

• The cost function for this target is multiplied by this weight

• Only relevant when there are multiple targets

• Epsilon
• The $$\epsilon$$ parameter in the relative difference measure

• Avoids infinities caused by dividing by zero

## Optimization Target Image Import

The spatial distribution of the target concentration or rate of change of concentration can be imported from a grayscale image in the same way as initial species concentrations.

• Minimum concentration
• The concentration corresponding to a black pixel in the image

• Maximum concentration
• The concentration corresponding to a white pixel in the image