mae as performance function for feedforward network in MATLAB2020b

net=feedforwardnet([5 5 5 5 5]);

%%

net.trainFcn='trainbr';

net.trainParam.epochs=12000;

net.divideFcn= 'dividerand'; % divide the data randomly

net.divideParam.trainRatio=0.85;

net.divideParam.valRatio=0.15;

% net.divideParam.testRatio=0.05;

net.performFcn= 'mae'; %newfcn

net.trainParam.mu=0.01; % Initial mu

net.trainParam.mu_dec=0.1; % mu decrease factor

net.trainParam.mu_inc=2.5; % mu increase factor

net.trainParam.mu_max=20e10; % Maximum mu

net.trainParam.max_fail=200;

net.trainParam.goal=1e-6;

%%

[net,tr]=train(net,S_tv1,e_tv','useParallel','yes','showResources','yes'); % ,'useParallel','yes','showResources','yes'

Why I have use mae, but the NNtool still show results of mse ? I am not sure whether matlab has

used the mae as the performance function.

Answer 1

[Sindhu Karri]

In the code using net.trainFunc as ‘trainbr’ MATLAB is generating warning message and is considering the default performance function (‘mse’) instead of ‘mae’

For detailed information refer to the Limitations section in the attached link:

https://www.mathworks.com/help/deeplearning/ref/trainbr.html