Optimizer class
Optimizer.RdThis class defines the optimization technique.
Public fields
id(`character(1)` The id of the objective.
print_trace(`logical(1)` Indicator whether to print the status of `$optimize()`.
Active bindings
lr(`numeric(1`) Step size of the algorithm.
archive(`data.table()`) Archive of all calls to `$evalStore`.
objective(`Objective`) The objective function.
x(`numeric()`) The numerical input vector used as starting point by `$optimize()`.
Methods
Method new()
Creates a new instance of this [R6][R6::R6Class] class.
Usage
Optimizer$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 `$optimize()`.
Method prepareUpdateForArchive()
Prepare updates for adding them to the archive.
Usage
Optimizer$prepareUpdateForArchive(
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`).