Re: Neural network - represention of age
- From: Russell.Martin@xxxxxxx
- Date: 4 Sep 2006 21:14:03 -0700
boltman wrote:
My original posted to comp.ai.neural-nets attracted the suggestion that this
forum might help.
I am experimenting with how to represent age in a back propagation feed
forward neural network that relates lifestyle factors to general health. I
have tried representing age as standard numeric input (e.g. 23), as age
banding (20-30 etc.), and as symbolic (young, middle aged etc). Has anyone
had any experience in trying out the various ways of representing age,
including perhaps using a fuzzy approach or introducing noise? If so, could
you please offer any advice.
Thanks
Bernie Shaw
I have not dealt with that particular question, but here's
a thought, FWIW. To the extent that death is arguably
related to some extent to poor health (death being the
poorest health) you might look at death rates as a function
of age and use that instead of age per se in some way as
an input. It's too late to think through how that might work
exactly...a problem left to the student, as they say. ;-)
Cheers,
Russell
.
- Follow-Ups:
- Re: Neural network - represention of age
- From: boltman
- Re: Neural network - represention of age
- References:
- Neural network - represention of age
- From: boltman
- Neural network - represention of age
- Prev by Date: Re: Correlation ratio = eta-squared, or eta?
- Next by Date: Re: Neural network - represention of age
- Previous by thread: Neural network - represention of age
- Next by thread: Re: Neural network - represention of age
- Index(es):
Relevant Pages
|