Click Here to Download NeuroSolutions Products Services Support Order Download Home

We have recently updated the NeuroDimension website. You are currently visiting an old page. Please click here to visit the updated version of our website.
NS - FAQ ...quick answers!



NeuroSolutions

Sections
General
Algorithms
Platforms & OS Issues
Network Components
Network Construction
Editing Networks
File Management
Running Networks

Contact Us

Download
Download the entire FAQ in Word 97(.doc) format.
If you don't have Word97....
Download the Word Viewer.

Network Components FAQ:

Q. What do Axons do?
A.
The Axon family sums all incoming vectors from multiple connections, and then applies a transfer function to the sum.

Q. What does the plain, blank Axon do?
A.
This Axon acts the same as all other Axons, except its transfer function is the identity function.

Q. What does a Full Synapse do?
A.
A Full Synapse takes its input and multiplies it by a matrix. If a delay is specified, the output is delayed by that many time steps.

Q. What does the straight Synapse do?
A.
Using the Straight synapse to connect two Axons is exactly the same as making a direct connection between them, except it gives the additional option of specifying a delay.

Q. What is that 2nd network that seems to lie on top of the main network?
A.
This is the backpropagation network. Every Axon and Synapse has a corresponding BackAxon and BackSynapse which attaches to the upper right corner of the corresponding forward component. Data flows forward from the input to the output through the forward propagation network. The criterion compares the output with the desired response, and computes the error. The error is then injected into the backpropagation network, and the data flows through this network, back towards the original input.

Q. In the control palette, what’s the difference between the clocks with one and two dials?
A.
The clocks with one dial are for use with static networks, where the complete forward activation and back-propagation cycle can be completed in the same time step. The clocks with two dials are dynamic controllers for use with dynamic networks (see the next question).

Q. When do I need to use the dynamic controllers?
A.
You need to use the dynamic controllers anytime your network has a feedback loop with an adaptable weight(s). This includes the Gamma and Laguarre Axons, which are components that have internal feedback loops with an adaptable weight. You also need the dynamic controllers whenever a component with delays is used at any point in the network other than the input layer. For example, you can use a static controller for a network with a TDNN Axon (tapped delay line) at the input layer, but you must use a dynamic controller if the TDNN Axon is in the hidden layer.

 


Products | Support | Order | Download | Search | Contact

Product questions? Contact info@nd.com Website questions? Contact webmaster@nd.com
Web Site Design and Implementation Copyright © 2002 NeuroDimension, Inc.