SearchBayes: SearchBayes Class

In five-dots/ml4e: Machine Learning for Everyone

Description

A class defines a bayesian search task specification.

Usage

new_search_bayes(options = NULL, keys = NULL, metric_name = NULL)

Arguments

options	A Options class object.
keys	A named character vector of keys.
metric_name	A character scalar of metric_name.

Value

A new SearchBayes class object

Super class

ml4e::Task -> SearchBayes

Active bindings

metric_name

A metric name.

is_new_grid

A logical wheather grid is new.

done

A logical if finished

error

A logical if error

result

A task result.

Methods

Public methods

• SearchBayes$new()

• SearchBayes$print()

• SearchBayes$do()

• SearchBayes$clone()

Inherited methods

• ml4e::Task$eval_call_args()

• ml4e::Task$print_task()

---

Method new()

Construct a new SearchBayes class object.

Usage

SearchBayes$new(options = NULL, keys = NULL, metric_name = NULL)

Arguments

options

A Options class object.

keys

A named character vector of keys.

metric_name

A character scalar of metric name. If NULL, the first metric_name in self$options$measure$metric_names is used.

Returns

A new SearchBayes class object

---

Method print()

Print object.

Usage

SearchBayes$print()

---

Method do()

Run a bayesian optimization. The result will be stored in self$result.

Usage

SearchBayes$do(
  init_grid_dt = NULL,
  init_points = 4L,
  n_iter = 5L,
  acq = "ucb",
  kappa = 2.576,
  eps = 0,
  kernel = list(type = "exponential", power = 2L)
)

Arguments

init_grid_dt

A data.frame of init_grid_dt.

init_points

A integer scalar of init_points.

n_iter

A integer scalar of n_iter.

acq

A character scalar of acq (aquisition function). Can be "ucb", "ei" or "poi".

kappa

A numeric scalar of kappa.

eps

A numeric scalar of eps.

kernel

A list of kernal parameters.

Returns

A data.frame of best parameters.

---

Method clone()

The objects of this class are cloneable with this method.

Usage

SearchBayes$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

five-dots/ml4e documentation built on June 19, 2020, 4:26 p.m.