This issue aims to add multiple cost evaluations to the optimiser and sampler classes. This would allow a list-like cost object to be passed to the optimiser/sampler for individual evaluation. The results are then stored in an optimisation/sampler result object for each cost. This issue relates to #238, specifically the N models and 1 dataset workflow.
This functionality is different from the MultiFittingProblem where multiple problems are combined into a single instance for cost evaluation, or the WeightedCost class where parameter values are weighted for a single cost function.
Motivation
This is a first step towards structured model comparison.
Possible implementation
Refactor the optimiser and sampler classes to accept a list of costs and individually optimise each. The Pints' ParallelEvaluator could be used for concurrent optimisation.
Feature description
This issue aims to add multiple cost evaluations to the optimiser and sampler classes. This would allow a list-like cost object to be passed to the optimiser/sampler for individual evaluation. The results are then stored in an optimisation/sampler result object for each cost. This issue relates to #238, specifically the
Nmodels and 1 dataset workflow.This functionality is different from the
MultiFittingProblemwhere multiple problems are combined into a single instance for cost evaluation, or theWeightedCostclass where parameter values are weighted for a single cost function.Motivation
This is a first step towards structured model comparison.
Possible implementation
Refactor the optimiser and sampler classes to accept a list of costs and individually optimise each. The Pints'
ParallelEvaluatorcould be used for concurrent optimisation.Additional context
No response