How to avoid declaration of neural network in an m file again and again?

I am using a feed forward neural network with keyword feedforwardnet in an m file.Whenever I run my m file neural network gives different results may be because of new weights and biases for each new declaration of neural network.I am not using init() function which initializes weights and biases each time; even then it is happening.It seems to happen because of new declaration of network each time m file is run.Please help me to avoid this so that I can obtain consistent results from neural network.

Answer 1

[Greg Heath]

rng(0)

% help rng %doc rng

Thank you for formally acceptingmy answer

Greg

[mittal54]

Thanks @greg I was wondering why rng(0) is used in most of the examples. Now I got it