get_optimized_parameters_nsga2r: Get optimized parameters from the output of...

In saeyslab/nichenetr: Modeling Intercellular Communication by Linking Ligands to Target Genes with NicheNet

Get optimized parameters from the output of run_nsga2R_cluster.

Description

get_optimized_parameters_nsga2 will take as input the output of run_nsga2R_cluster and extract the optimal parameter values, either from the best solution at the end of the generations or the best solution across all generations.

Usage

get_optimized_parameters_nsga2(result_nsga2r, source_names, search_all_iterations = FALSE, top_n = NULL, summarise_weights = TRUE)

Arguments

result_nsga2r	The output of run_nsga2R_cluster.
source_names	Character vector containing the names of the data sources.
search_all_iterations	Logical indicating whether the best solution across all generations should be considered (TRUE) or only the best solution at the end of the generations (FALSE).
top_n	If search_all_iterations=TRUE, this indicates how many of the best solutions should be considered.
summarise_weights	If search_all_iterations=TRUE, a logical indicating whether the weights should be summarised by taking the mean and median.

Value

A list containing two dataframes, the optimal data source weights and the optimal hyperparameters.

Examples

## Not run: 
results <- run_nsga2R_cluster(model_evaluation_optimization_nsga2r, varNo=n_param, objDim=n_obj,
lowerBounds=lower_bounds, upperBounds=upper_bounds, popSize = 360, tourSize = 2, generations = 15, ncores = 8)

# Get the best solution at the end of the generations
optimized_parameters <- get_optimized_parameters_nsga2(results, source_names, search_all_iterations = FALSE, top_n = NULL, summarise_weights = TRUE)

# Get the best solution across all generations, consider top 25 solutions and summarise weights
optimized_parameters <- get_optimized_parameters_nsga2(results, source_names, search_all_iterations = TRUE, top_n = 25, summarise_weights = TRUE)

## End(Not run)

saeyslab/nichenetr documentation built on Sept. 7, 2024, 10:27 a.m.