grid_search_1d {structToolbox}R Documentation

grid_search_1d class

Description

Carries out a grid search for a single parameter to try and identify the 'best' value for the parameter based on the input metric.

Usage

grid_search_1d(
  param_to_optimise,
  search_values,
  model_index,
  factor_name,
  max_min = "min",
  ...
)

Arguments

param_to_optimise

The name of an input parameter of the model the optimise

search_values

A vector of values to search for the optimum

model_index

A number indicating which step of a model_seq is to be optimised

factor_name

The sample_meta column name to use

max_min

'A string 'max' or 'min' to indicate whether to maximise or minimise the metric

...

additional slots and values passed to struct_class

Value

struct object

Examples

D = MTBLS79_DatasetExperiment()
# some preprocessing
M = pqn_norm(qc_label='QC',factor_name='class') +
    knn_impute() +
    glog_transform(qc_label='QC',factor_name='class') +
    filter_smeta(factor_name='class',levels='QC',mode='exclude')
M=model_apply(M,D)
D=predicted(M)

# reduce number of features for this example
D=D[,1:10]

# optmise number of components for PLS model
I = grid_search_1d(param_to_optimise='number_components',search_values=1:5,
        model_index=2,factor_name='class') *
        (mean_centre()+PLSDA(factor_name='class'))
I = run(I,D,balanced_accuracy())


[Package structToolbox version 1.0.1 Index]