Error in the configuration for a Genetic Algorithm - Integral Values output

Question

0 个投票

Im trying to find the correct number of neurons that go into each of the 3 Hidden Layers in an NN through a genetic algorithm, that gives out integral values for number of neurons in each layer.

My function takes in the arguments for number of neurons in each layer, and outputs the RMSE for that configuration. Heres the function I wanna minimize.

function output_rmse = CalculateBestConfiguration(NeuronsIn_HL_1,NeuronsIn_HL_2,NeuronsIn_HL_3)
%% Create Pipeline
trainer1 = MLP_pipeline;
%% Generate Data
[X_data , Y_data, q1_set, ~, ~,~,~,~,~] = PlotWorkspace(10,7,5);
[trainInd,testInd] = dividerand(length(q1_set),80,20);
%% Feed Training Data
trainer1.input_training = [X_data(trainInd);Y_data(trainInd)];
trainer1.output_training = q1_set(trainInd);
%% Train the Perceptron Model
trainer1.fit_MLP([NeuronsIn_HL_1,NeuronsIn_HL_2,NeuronsIn_HL_3],'trainbr');
trainer1.train_MLP()
%% Feed Test Data
trainer1.input_test = [X_data(testInd);Y_data(testInd)];
trainer1.output_test = q1_set(testInd);
%% Predict Results
theta_predicted1 = trainer1.predictor();
%% Evaluate Accuracy
error_values1 = trainer1.generate_error(theta_predicted1);
output_rmse = sqrt(mean(error_values1.^2));
end

Theres something wrong with my configuration for the GA. I am not able to find it.

opt = optimoptions('ga');
% opt.InitialPopulationMatrix = ones(30,3);
opt.MaxTime = 4*60*60;
opt.MigrationFraction = 1.0000e-06;
% opt.MutationFunction= 'mutationgaussian';
opt.PopulationSize = 30;
opt.FunctionTolerance = 1e-4;
opt.MaxGenerations = 500;
fun = @CalculateBestConfiguration;
IntCon = 1:3;
lb = [1,1,1];
ub = [21,21,21];
A = [3 3 3];
b = 180;
nvars = 3;
nonlcon = [];
Solution = ga(fun,nvars,A,b,[],[],lb,ub,nonlcon,IntCon,opt);

Here's my error:

Not enough input arguments.
Error in CalculateBestConfiguration (line 11)
trainer1.fit_MLP([NeuronsIn_HL_1,NeuronsIn_HL_2,NeuronsIn_HL_3],'trainbr');
Error in createAnonymousFcn>@(x)fcn(x,FcnArgs{:}) (line 11)
fcn_handle = @(x) fcn(x,FcnArgs{:});
Error in fcnvectorizer (line 13)
            y(i,:) = feval(fun,(pop(i,:)));
Error in gaminlppenaltyfcn
Error in gapenalty
Error in makeState (line 64)
            Score = FitnessFcn(state.Population(initScoreProvided+1:end,:));
Error in galincon (line 17)
state = makeState(GenomeLength,FitnessFcn,Iterate,output.problemtype,options);
Error in gapenalty
Error in gaminlp
Error in ga (line 393)
    [x,fval,exitFlag,output,population,scores] = gaminlp(FitnessFcn,nvars, ...
Error in untitled (line 18)
Solution = ga(fun,nvars,A,b,[],[],lb,ub,nonlcon,IntCon,opt);
Caused by:
    Failure in user-supplied fitness function evaluation. GA cannot continue.
    Failure in initial user-supplied fitness function evaluation. GA cannot continue.
 

2 个评论
显示无隐藏无

Walter Roberson 2019-4-10

What are MLP_Pipeline and fit_MLP ? They are not part of any toolbox I can find.

Aniketh Ramesh 2019-4-10

在 MATLAB Online 中打开

They are wrappers I have written for the fit and train functions. Here:

classdef MLP_pipeline < handle
    properties
        input_training
        input_test
        output_training
        output_test
        fitted_net
        trained_net
    end
    methods
        function obj = fit_MLP(obj,neural_net_param,training_function)
           obj.fitted_net = fitnet(neural_net_param,training_function); 
        end
        function obj = train_MLP(obj)
           obaj.trained_net = train(obj.fitted_net,obj.input_training,obj.output_training); 
        end
        function predicted_output = predictor(obj)
            predicted_output = obj.trained_net(obj.input_test);
        end
        function error_values = generate_error(obj,predicted_output)
            error_values = gsubtract(obj.output_test,predicted_output);
        end
        function performanceValue = generatePerformance(obj,predicted_output)
            performanceValue = perform(obj.trained_net,obj.output_test,predicted_output);
        end
    end
end