Bayesopt result is "No feasible points were found." in classifier optimization

Hello everyone,

I need to tune a random forest in a classification task and I am following this guide from matlab documentation that does the same but for regression.

https://uk.mathworks.com/help/stats/tune-random-forest-using-quantile-error-and-bayesian-optimization.html

I modified the code to optimize a classifier, but I I'm struggling in understanding why bayesopt can't find any feasible point. This is an example of what I get:

|=====================================================================================================|
| Iter | Eval   | Objective   | Objective   | BestSoFar   | BestSoFar   |        minLS |       numPTS |
|      | result |             | runtime     | (observed)  | (estim.)    |              |              |
|=====================================================================================================|
|    1 | Error  |         NaN |     0.19415 |         NaN |         NaN |           20 |            3 |
|    2 | Error  |         NaN |     0.20093 |         NaN |         NaN |            2 |            1 |
|    3 | Error  |         NaN |      0.2076 |         NaN |         NaN |            2 |            4 |
|    4 | Error  |         NaN |     0.19925 |         NaN |         NaN |           17 |            6 |
|    5 | Error  |         NaN |     0.19505 |         NaN |         NaN |           13 |            2 |

__________________________________________________________
Optimization completed.
MaxObjectiveEvaluations of 5 reached.
Total function evaluations: 5
Total elapsed time: 1.7825 seconds.
Total objective function evaluation time: 0.99699
No feasible points were found.
    

I also add my code in case it could be helpful:

function  bestHyperparameters = RF(trainingData,predictorNames)
rng('default'); % For reproducibility

% Extract predictors and response
inputTable = trainingData;
predictors = inputTable(:, predictorNames);
response = inputTable.Class;

% Set hyperparameters
maxMinLS = 20;
minLS = optimizableVariable('minLS',[1,maxMinLS],'Type','integer');
numPTS = optimizableVariable('numPTS',[1,size(predictors,2)-1],'Type','integer');
hyperparametersRF = [minLS; numPTS];

% obj. function
    function oobErr = oobErrRFM(params,X,response)        
        randomForest = TreeBagger(30,X,response,'Method','classification',...
            'OOBPrediction','on','MinLeafSize',params.minLS,...
            'NumPredictorstoSample',params.numPTS);
        oobErr = oobError(randomForest);
    end

% Optimization
results = bayesopt(@(params)oobErrRFM(params,predictors,response),hyperparametersRF,...
    'Verbose',1,'MaxObjectiveEvaluations',5);

bestOOBErr = results.MinObjective;
bestHyperparameters = results.XAtMinObjective;
end

I hope someone can help me! Thank you in advance,

Marta

[Stephan]

remove your comment and make an answer to this question instead. then accept your answer. this way people know the issue is solved successfully.

Answer 1

[marta prati]

After posting the question, I found the error...

If it could be helpful to someone, the code works just by replacing oobErr = oobError(randomForest) with oobErr = oobError(randomForest, 'Mode','ensemble').

Cheers!