Page 1 of 1

How to map the parameters with good and bad effect in input?

PostPosted: Sun Jan 17, 2010 3:38 am
by Edrisfm
Hello
I use MLP for soccer prediction. There are some input properties. It seems that some of them have good effect on result(like: shoots on target count) and some of them have bad effect(like bad passes count).

Now, my question is in mapping data for input, I should map both of properties in a same approach(NN modifies weights correctly) or I should map data with negative approach(for example: 1's complement)

Re: How to map the parameters with good and bad effect in input?

PostPosted: Mon Jan 18, 2010 11:19 am
by andrew.kirillov
Hello,

Hard to say for sure without learning the problem deeply. But why don't we let neural network to figure it out? As you've mentioned, during learning phase it will update its weights on its own.