Contents
Description of demo_regression_MLP
Multi-layer perceptron regression with sigmoid and hyperbolic tangent activiation functions
clear all close all generateData_volatile
usage of MLP regression with sigmoid activation function
options_mlp1.nHidden= [8,8,8,8]; options_mlp1.activFunc = 'sig'; [model_mlp1] = ml_regression_MLP(Xtrain,ytrain,options_mlp1); yhat_mlp1 = model_mlp1.predict(model_mlp1, Xtest); testError_mlp1 = mean(abs(yhat_mlp1 - ytest)); fprintf('Averaged absolute test error with %s is: %.3f\n', model_mlp1.name, testError_mlp1);
Averaged absolute test error with Multi-layer Perceptron with Sigmoid/Logistic Activation Function is: 0.408
usage of MLP regression with tanh activation function
options_mlp2.nHidden= [8,8,8,8]; options_mlp2.activFunc = 'tanh'; [model_mlp2] = ml_regression_MLP(Xtrain,ytrain,options_mlp2); yhat_mlp2 = model_mlp2.predict(model_mlp2, Xtest); testError_mlp2 = mean(abs(yhat_mlp2 - ytest)); fprintf('Averaged absolute test error with %s is: %.3f\n', model_mlp2.name, testError_mlp2);
Averaged absolute test error with Multi-layer Perceptron with tanh Activation Function is: 0.248
plotRegression1D(Xtrain, ytrain, model_mlp1, model_mlp2);
