Variable Types

Algorithms in optiseek can accept several variable types in addition to just continuous values. Internally, the optimizers treat all variables as continuous numerical dimensions in the search space; however, they are converted back to the appropriate user-specified format when passed to the objective function. Currently, supported variable types are continuous (float), integer, categorical/ordinal, and boolean.

Each variable type has its own class. When defining variables and search space bounds for an algorithm, the user passes a list of variable classes into the var_list argument of the algorithm class.

Variable Type Class Instantiation
Floats/Continuous optiseek.variables.var_float(var_name, bounds, log_scale=False)
Integers optiseek.variables.var_int(var_name, bounds, log_scale=False)
Categorical optiseek.variables.var_categorical(var_name, choices)
Boolean (True/False) optiseek.variables.var_bool(var_name)

Parameters

Parameter Description
var_name : string Name of the variable. This will be used to track output and label the
stored results if applicable.
bounds : list of floats/int A list containing a lower and upper bound for the search space of the
variable in the format [lower, upper]. The values can be integers/floats
for var_float and must be integers for var_int.
choices : list Contains a list of choices to be passed as an argument for the objective
function. The list items may be any type that the objective function
can accept.
log_scale : boolean A flag to toggle on logarithmic scaling for a specified variable. For
example, for a continuous variable with search bounds [1, 100], the
midpoint without log scaling is 50.5. If log_scale was set to True, the
distance in the search space from 1 to 10 would be equivalent to that
from 10 to 100. In order to use log_scale, the search space for that
variable must be positive.

Example

from optiseek.variables import *

# defining a variable list to be passed to an algorithm
var_list = [
    var_float('x_float', [-10.5, 10.5]),
    var_int('x_int', [-2, 5]),
    var_categorical('x_cat', ['small', 'medium', 'large']),
    var_bool('x_bool')
]