Skip to contents

Creates settings for the package's default non-temporal transformer model.

Usage

setDefaultTransformer(
  estimatorSettings = setEstimator(learningRate = "auto", weightDecay = 1e-04, batchSize
    = 512, epochs = 10, seed = NULL, device = "cpu")
)

Arguments

estimatorSettings

Estimator settings created by setEstimator().

Value

A modelSettings object for use with PatientLevelPrediction.

Details

The architecture and default hyperparameters are based on Gorishniy et al. (2021).

Examples

transformerSettings <- setDefaultTransformer()
transformerSettings$param[[1]]$numBlocks
#> [1] 3