gridSearch: Grid Search

Description

Given a set of possible hyperparameter values, the function trains models with all the possible combinations of hyperparameters.

Usage

gridSearch(model, hypers, metric, test = NULL, bg4test = NULL,
  env = NULL, parallel = FALSE, save_models = TRUE, seed = NULL)

Arguments

model

'>SDMmodel or code'>SDMmodelCV object.

hypers

named list containing the values of the hyperparameters that should be tuned, see details.

metric

character. The metric used to evaluate the models, possible values are: "auc", "tss" and "aicc".

test

code'>SWD object. Test dataset used to evaluate the model, not used with aicc and code'>SDMmodelCV objects, default is NULL.

bg4test

Deprecated.

env

stack containing the environmental variables, used only with "aicc", default is NULL.

parallel

logical, if TRUE it uses parallel computation, default is FALSE.

save_models

logical, if FALSE the models are not saved and the output contains only a data frame with the metric values for each hyperparameter combination. Default is TRUE, set it to FALSE when there are many combinations to avoid R crashing for memory overload.

seed

Deprecated.

Value

code'>SDMtune object.

Details

To know which hyperparameters can be tune you can use the output of the function get_tunable_args. Hyperparameters not included in the hypers argument take the value that they have in the passed model. Parallel computation increases the speed only for large datasets due to the time necessary to create the cluster.

Examples

Run this code

# NOT RUN {
# Acquire environmental variables
files <- list.files(path = file.path(system.file(package = "dismo"), "ex"),
                    pattern = "grd", full.names = TRUE)
predictors <- raster::stack(files)

# Prepare presence and background locations
p_coords <- virtualSp$presence
bg_coords <- virtualSp$background

# Create SWD object
data <- prepareSWD(species = "Virtual species", p = p_coords, a = bg_coords,
                   env = predictors, categorical = "biome")

# Split presence locations in training (80%) and testing (20%) datasets
datasets <- trainValTest(data, test = 0.2, only_presence = TRUE)
train <- datasets[[1]]
test <- datasets[[2]]

# Train a model
model <- train(method = "Maxnet", data = train, fc = "l")

# Define the hyperparameters to test
h <- list(reg = 1:2, fc = c("lqp", "lqph"))

# Run the function using the AUC as metric
output <- gridSearch(model, hypers = h, metric = "auc", test = test)
output@results
output@models
# Order rusults by highest test AUC
head(output@results[order(-output@results$test_AUC), ])

# Run the function using the AICc as metric and without saving the trained
# models, helpful when numerous hyperparameters are tested to avoid memory
# problems
output <- gridSearch(model, hypers = h, metric = "aicc", env = predictors,
                     save_models = FALSE)
output@results
# }