How to get the same output for the same test example provided to the neural network as explained below?

Hi everyone,

I'm building a property-rental prediction model using 2-layer BackPropagation neural network. Everything went well. But the net produces different output every time I apply it to the same test example. I'd like to know how can I get consistent results when applying the network on the same test data (using sim function)? I know that it has something to do with the network weights being randomly initialized. However, there must be a way to either initialize the weights to some values or maybe to use another strategy.

Thank you!

Iulia

Answer 1

[Greg Heath]

Intialize the random number generator to a specified state before the weights are initialized. For example, rng(0) or rng(1492)

Hope this helps.

Thank you for formally accepting my answer

Greg

P.S. Search rng(0) greg

[Iulia]

Thanks Greg. Yes it worked just by adding rand('state',0) But I still wonder if this way is the best. I mean, does the random initialization of weights has better effect on training than static initialization?

Thanks!