Skip to contents

Optimizer class

Optimizer.Rd

This class defines the optimization technique.

Public fields

id

(character(1))
The id of the object.

print_trace

(logical(1) Indicator whether to print the status of $optimize().

Active bindings

lr

(numeric(1) Step size of the algorithm.

archive

(data.table()) Archive of all calls to $eval_store.

objective

(Objective) The objective function.

x

(numeric()) The numerical input vector used as starting point by $optimize().

Methods

Public methods

Method new()

Creates a new instance of this R6 class.

Usage

Optimizer$new(objective, x_start, id = NULL, print_trace = TRUE)

Arguments

objective

(Objective)
The objective to optimize.

x_start

(numeric())
Start value of the optimization. Note, after the first call of $optimize() the last value is used to continue optimization. Get this value with $x.

id

(character(1))
Id of the object.

print_trace

(logical(1))
Indicator whether to print the status of $optimize().

Method prepare_update_for_archive()

Prepare updates for adding them to the archive.

Usage

Optimizer$prepare_update_for_archive(
  x_out,
  x_in,
  update,
  fval_out,
  fval_in,
  lr,
  step_size,
  objective,
  step,
  ...
)

Arguments

x_out

(numeric()) The new proposed point by the optimizer.

x_in

(numeric()) The old input value which is updated to x_out.

update

(numeric()) The update from x_in to x_out.

fval_out

(numeric(1)) The objective value objetive$eval(x_out).

fval_in

(numeric(1)) The objective value objetive$eval(x_in).

lr

(numeric(1)) The learning rate used to multiply update with.

step_size

(numeric(1)) The step_size used to multiply lr * update with.

objective

(Objective) The objective used by $optimize().

step

(integer(1)) The step or iteration.

...

Additional objects added to the archive (e.g. momentum).

Returns

data.table() of the input arguments.

Method update_archive()

Add points to the archive.

Usage

Optimizer$update_archive(ain)

Arguments

ain

data.table() with names "x_out", "x_in", "update", "fval_out", "fval_in", "lr", "objective", and "step".

Method set_x()

Set the current input vector used as start point of $optimize().

Usage

Optimizer$set_x(x)

Arguments

x

(numeric()) Input vector.

Method clone()

The objects of this class are cloneable with this method.

Usage

Optimizer$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.