Greetings
from NeuroDimension!
The World Leader in Neural Network Software
This issue of the NeuroDimension newsletter highlights the latest
developments at NeuroDimension, as well as providing insights on improving your neural networks
with genetic optimization.
In this issue you’ll find:
What’s New
and News?
* New Neural Network Course
Announced
Note: You are
receiving this newsletter because you requested to stay informed concerning new
developments at NeuroDimension. If you would like to stop receiving these
newsletters, please see the bottom of this newsletter for removal instructions.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
The next neural network course has been scheduled for April
28 – May 2, 2002 at the Doubletree Hotel and Conference Center located in
Gainesville, Florida. A 10% discount is available for early registration, so be
sure to register today!
Our course format allows both novice and advanced users to
find a suitable course. Offered courses include: "Introduction to
NeuroSolutions", "Fundamentals of Neural Networks and
NeuroSolutions", and "Advanced NeuroSolutions". The courses
include a copy of our interactive book, Neural and Adaptive Systems:
Fundamentals Through Simulations.
For details on this new offering, or to sign-up from the
Internet, see http://www.neurosolutions.com/products/course/april_2003.html
For general ND course information, see http://www.neurosolutions.com/products/course/
For more information and samples of the interactive book,
see http://www.neurosolutions.com/products/nsbook/
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
NeuroSolutions Tip Box
Genetic optimization is a feature of NeuroSolutions
Users level and above. It allows you to optimize virtually any parameter in a
neural network to produce the lowest error. For example, the number of hidden
units, the memory taps, the learning rates, and the input selection can all be
optimized to improve the network performance.
A common problem that people encounter when using genetic
optimization is that many people choose to optimize all of the parameters when
first specifying a neural network to solve a particular problem. This approach
can often result in a lot of wasted processor time, since the genetic
optimization can take many hours to train and the problem may not be properly
defined to begin with.
Developing an Optimized Model
Here is a set of steps that can be used as a basis for
developing an optimized model while making sure the underlying approach is
valid to begin with.
1. First configure the neural network using either the
NeuralExpert or NeuralBuilder. In either case, do not choose genetic
optimization.
2. Train and test the neural network to make sure that it
produces reasonable results. If the results are not reasonable, then you need
to evaluate the input, output and desired data to try to determine why. Genetic
optimization will not fix a broken network – it will only improve on one that
works.
3. If you have several inputs into the network and you are
not sure if all are needed, the next step might be to use genetic optimization
to exclude those unnecessary inputs. You can add genetic optimization to an
existing breadboard by stamping the GeneticControl component on top of the
BackStaticControl (the red dials in the upper left corner of the breadboard).
To select the inputs to optimize, go to the inspector of the input File
component, select “Customize” and check the GA checkboxes next to the column
labels. Make any necessary changes to the parameters within the GeneticControl
inspector and run the network.
4. Once you have determined the set of inputs to use, you
can then try to optimize the number of hidden processing elements (PEs) and/or
input taps. This is specified within the “Genetic” tab of the Axons’ inspector
as the “Rows” and “Taps” parameters. The learning rates often do not have a
dramatic effect on the final result, but these can also be optimized at the
same time by selecting the “Step Size” parameter within the “Genetic” tab of
the Gradient Search components.
Setting Up Genetic Optimization
You do not have to set up the genetic optimization by hand
-- the NeuralBuilder and the NeuralExpert have this capability. With the
NeuralExpert, there are three settings – low, medium and high. Selecting the
“low” level of genetic optimization allows the genetic optimizer to manipulate
the learning rates. The “medium” level sets the learning rates and number of
PEs in each layer. The “high” level allows the genetic optimizer to set most of
the parameters in the neural network.
The NeuralBuilder gives you a little more flexibility in
that you can specify exactly which parameters you want optimized. However, the
NeuralExpert has the advantage that it will allow you to modify an existing
breadboard to add genetic optimization. If you want to add genetic optimization
to a neural network built with the NeuralBuilder, you need to re-define the
network within the NeuralBuilder, perform the actions described in steps 3 and
4 above, or use NeuroSolutions for Excel.
Using Genetic Optimization from NeuroSolutions for Excel
The easiest way to optimize the parameters of a
NeuroSolutions neural network is to use NeuroSolutions for Excel. Simply select
“Train” then “Train Genetic…” from the NeuroSolutions menu of Excel. This
allows you to easily specify which parameters you want optimized for the
training run, without having to manipulate the components within NeuroSolutions
or re-build the network.
=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=~=
We appreciate your feedback! Please send us your comments or
suggestions concerning this newsletter, our web site, or part of the NeuroDimension
product line. Write to us at: feedback@nd.com
Have questions about NeuroDimension products or training
services? Send your questions to: info@nd.com
Some mail readers may display an attachment at the bottom of
this newsletter. This is typically caused when the mail reader is set to
display text only. The original, formatted version is sometimes included as an
attachment.
This issue and previous issues of this newsletter are
available on the NeuroDimension web site at: http://www.neurosolutions.com/newsletters.html
If you would prefer not to receive these newsletters or
subsequent product updates from NeuroDimension, please reply to this letter
with the subject heading changed to the word REMOVE.
Thank you again for your support of NeuroDimension products!