Bayesian Optimization

Reference

#

Honestly, I’m not familiar with BayesianOPT, the opinions mentioned stem from the below. 👇

• 【机器学习】一文看懂贝叶斯优化/Bayesian Optimization

• 一文详解贝叶斯优化（Bayesian Optimization）原理

• 贝叶斯优化(BayesianOptimization)

• 超参数优—贝叶斯优化及其改进（PBT优化）

• 贝叶斯优化 (Bayesian Optimization)

• MATLAB Offical Document

Advantages & Algorithm Principle

#

Here we are going to talk about the advantages & algorithm principle of BayesianOPT. If you only want to konw how to use it, you can read the #Advantage section, then go to the MATLAB Practice

Advantages

#

Algorithm Principle

#

MATLAB Practice

#

Well, we can put Bayesian Optimization into practice even though we don’t understand it, using the predefined function of MATLAB, the bayesopt. Here is the official guidance of the function: bayesopt

Final code display

#

% define the obj function
function y = objectiveFcn(x)
    y = (1 - x.x1)^2 + 100 * (x.x2 - x.x1^2)^2;
end

% define the variables
vars = [optimizableVariable('x1', [-2, 2])
        optimizableVariable('x2', [-2, 2])];

% conduce the optimizer
results = bayesopt(@objectiveFcn, vars, ...
                   'AcquisitionFunctionName', 'expected-improvement-plus', ...
                   'MaxObjectiveEvaluations', 30, ...
                   'IsObjectiveDeterministic', true, ...
                   'Verbose', 1);

% get result
bestPoint = results.XAtMinObjective;
bestObjective = results.MinObjective;

% result output
fprintf('最优解 x1: %.4f, x2: %.4f\n', bestPoint.x1, bestPoint.x2);
fprintf('最优目标值: %.4f\n', bestObjective);

Parameters Explaination

#

Want more detailed information? Refer to the Offical document: bayesopt. It’s more completed and with amount of examples.