soo_base Overview
This page introduces the soo_base project.
This project optimizes the single-objective benchmark function Sphere[11].
optimization.yaml (soo_base project)
The contents of optimization.yaml are shown below.
The evaluator is the sphere implementation of the Sphere function, with dim, the number of design variables (the dimension of the optimization problem), set to 30.
The optimizer is the single-objective optimization algorithm cmaes with a fixed random seed. The design variables are constrained to the range [-5.12, 5.12] by setting bounds.
evaluator:
name: sphere
kwargs:
dim: 30
optimizer:
name: cmaes
kwargs:
bounds: [-5.12, 5.12]
seed: 42
This example performs 100 optimization iterations.
num_iteration: 100
(Output-related settings are omitted.)
Optimization Example
The GUI after completing 100 optimization iterations is shown below. The function value gradually converges to 0.
