Scikit 优化和离散变量

Question

scikit-optimize has two classes for definind the variable type: Integer and Categorical. The issue is that integer space class handles only true integers, while categorical class does not interpret the values as numbers (does not define distance betwee points - as per ducumentation).

Is there a way to define a variable as a list of discrete values, like for example [1, 2, 4, 6, 8, 10, 20, 30, 40]?

Answer 1

skopt.space.Space offers three types that a search space dimension can take: Real , Integer or Categorical .

You can define a 3D space that contains only integer dimensions the following way:

from skopt.space import Space, Integer
space = Space(Integer(1, 2, 4), Integer(2, 4, 6))