Page 1 of 1

Nearal Network gives different results on repeat

PostPosted: Tue May 17, 2016 11:36 am
by Lawrence
Hi,

I am using a version of the code in TimeSeries demo.

I notice that I get different solutions(slightly) when I run the demo several times.

Is there a hidden random element in compute that might cause this?

Re: Nearal Network gives different results on repeat

PostPosted: Tue May 17, 2016 3:03 pm
by andrew.kirillov
Hello,

Neurons' weights are randomly initialized. If you need same results, you need to set Neuron.RandGenerator every time before running a test with a new network.

Re: Nearal Network gives different results on repeat

PostPosted: Tue May 17, 2016 3:41 pm
by Lawrence
That's perfect.

Thank you