grid_to_nano: Convert Grid to Nano Object

In Nanoputian628/nano: Data Visualisation and Model Selection

Convert Grid to Nano Object

Description

Converts a grid in a nano object to a separate nano object.

Usage

grid_to_nano(
  nano,
  grids_no = nano$n_model,
  n_top_model = length(nano$grid[[grids_no]]@model_ids)
)

Arguments

nano	nano object containing the grid to be converted to a new nano object.
grids_no	a numeric. Vector containing positions of grids to be converted to a nano object.
n_top_model	a numeric. Vector of the n top number of models to select from each grid.

Details

Converts a specified grid in a nano object to a new nano object where a new slot is created for each model in the specified grid. The purpose of this function is to be used when performing hyper-parameter tuning (using the nano_grid function). When hyper- parameter tunning is performed, a grid of models is created, however, only a single model will be stored in the nano object. The diagnostic functions (e.g. nano_pdp, nano_ice, etc) are only able to analyse models in different slots of the nano object. Hence, to compare the models in the grid with each other, this function should be run to create a new nano object with a slot of each model. Then the usual diagnostic functions can be run to compare the models. Later, if you wish to select a different model from the grid (by default, the model with the best metric will be stored in the nano object), then the switch_model function can be run.

Value

a nano object

Examples

## Not run: 
if(interactive()){
 library(h2o)
 library(nano)
 
 h2o.init()
 
 # import dataset
 data(property_prices)
 
 # set the response and predictors
 response <- "sale_price"
 var <- setdiff(colnames(property_prices), response)
 
 # hyper-parameter tuning
 nano <- nano_grid(data         = property_prices, 
                   response     = response, 
                   algo         = "drf",
                   hyper_params = list(ntrees = 1:3))
                   
 # convert created grid to new nano object with 3 slots
 nano_1 <- grid_to_nano(nano, 1)
 
 }

## End(Not run)

Nanoputian628/nano documentation built on Oct. 30, 2023, 3:28 p.m.