| Computers Forum Index » Computer Artificial Intelligence - Neural Nets » Inputs direct to second ("hidden") layer as well as... |
|
Page 1 of 1 |
|
| Author |
Message |
| ... |
 Posted: Mon Oct 19, 2009 12:55 pm |
|
|
|
Guest
|
Does anyone know of any work where (some) inputs go directly to the
hidden
layer as well as (other) inputs going into the first ("input") layer?
(Perhaps to
help control concept learning in the hidden layer.) I'm assuming a
feedforward
net. |
|
|
| Back to top |
|
|
|
| Stephen Wolstenholme... |
| Posted: Mon Oct 19, 2009 5:19 pm |
|
|
|
Guest
|
On Mon, 19 Oct 2009 05:55:43 -0700 (PDT), jonesrob at (no spam) emporia.edu wrote:
Quote: Does anyone know of any work where (some) inputs go directly to the
hidden
layer as well as (other) inputs going into the first ("input") layer?
(Perhaps to
help control concept learning in the hidden layer.) I'm assuming a
feedforward
net.
I added that type of configuration to the network editor in my
EasyNN-plus application a few years ago. So far I have yet to see a
problem when that type of skip configuration is needed. It can make
learning faster but that's about it.
Steve
--
Neural Planner Software Ltd www.NPSL1.com |
|
|
| Back to top |
|
|
|
| Greg Heath... |
| Posted: Tue Oct 20, 2009 1:03 pm |
|
|
|
Guest
|
On Oct 19, 8:55 am, jones... at (no spam) emporia.edu wrote:
Quote: Does anyone know of any work where (some) inputs go directly to the
hidden
layer as well as (other) inputs going into the first ("input") layer?
(Perhaps to
help control concept learning in the hidden layer.) I'm assuming a
feedforward
net.
Searching the CANN archives in Google Groups
skip -jump layer 78 hits
-skip jump layer 54 hits
skip jump layer 8 hits
Hope this helps.
Greg |
|
|
| Back to top |
|
|
|
|
