How to combat the problem of overfitting in a Matlab Neural Network

I constructed my own neural network with matlab but it is a bit overfitted. Online I red that with the function net = feedforwardnet(20,'trainbr'); but with it the overfitting is not much better. The data is already distributed randomly over the samples. I also tried to combine the forecasting results of more networks.

My goal is to change the minimization goal so that it includes a term of the squared parameters calculated by the NN