Custom cost function in deep learning toolbox

I want to use a custom cost function in neural network that minimizes the mean square error between . I followed the steps mentioned here:

https://www.mathworks.com/matlabcentral/answers/461848-custom-performance-function-for-neural-networks-toolbox-using-mse-package

I created a file myfcn.m and +myfcn package.

But when I run thte code with custom cost function

net.performFcn = 'myfcn'

I get the following error:

Error using network/subsasgn>network_subsasgn (line 555)
"performFcn" cannot be set to non-existing function "myfcn".

Error in network/subsasgn (line 14)
net = network_subsasgn(net,subscripts,v,netname);

Error in SDR_exp_customfcn (line 79)
net.performFcn = 'myfcn';

I couldn't figure out what the issue is.

Answer 1

[Abbott Von]

https://www.mathworks.com/matlabcentral/answers/461848-custom-performance-function-for-neural-networks-toolbox-using-mse-package

Well， maybe this tutorial can help you. The key is that line "Defining custom functions with this approach works only with the non-MEX version of the Neural Network code, so it is necessary to call 'train' using the 'nn7' option."

net = train(net,x,t,nn7);

y = net(x,nn7);

Answer 2

[Abbott Von]

copyfile([matlabroot '\toolbox\nnet\nnet\nnperformance\mse.m'], 'myfcn.m');

copyfile([matlabroot '\toolbox\nnet\nnet\nnperformance\+mse'], '+myfcn');

And pull the necessary functions into your current working directories.