Innovation Award
June 2004
Winner |
Neural Networking is a computing science field that is not so well known among most software developers, but it can help solving problems that are difficult to solve with conventional computing methods.
Neural Networking consists of emulating the way a brain works by the means of networks of interconnected neurons. These neurons have inputs and outputs. They receive input signals that are combined according to some internal factors to define their output signals.
The neural networks need to be trained to perform some useful tasks. Usually these tasks consist of perceiving patterns in the input signals of their neurons. This is why they are also known as perceptrons.
Training consist of submitting a reasonably large sample of subjects that need to be recognized. Iteratively the values of internal factors of each neuron are adjusted until they output the desired response with an acceptable error rate.
The most common applications of neural networks are usually recognizing patterns that otherwise would be hard to distinguish by common computing methods, like for instance, recognizing handwritten characters in a scanned image.
This class provides an implementation of a perceptron. It comes with an example application of classifying users by type of profile based on the answers that they provide to questions presented in a Web form.
Since the class provides a generic implementation of a perceptron, there could be many other examples of application for this class, possibly only limited by the imagination of the developers that understand its potential.
Manuel Lemos |
This class implements a model of the Percetron (Artificial Neural Networks - ANN) that can be trained to recognize patterns in its inputs.
This is similar to the algorithm used on palmtops to recognize words written on its pen pad.
The example that comes with this class demonstrates how it can be used to find people that match the profile an inquiring user that fills a form with questions about personal preferences.
First the perceptron is trained with answers of several people. Then the answers of an inquiring user are converted into values that are fed to the perceptron in order to find profile matches.
| Ratings | Utility |
Consistency |
Documentation |
Examples |
Tests |
Videos |
Overall |
Rank |
| All time: |
Sufficient (77.8%) |
Sufficient (69.4%) |
Not sure (56.3%) |
Sufficient (69.4%) |
- |
- |
Not sure (58.4%) |
813 |
| Month: |
Not yet rated by the users |
No application links were specified for this class.
If you know an application of this package, send a message to the
author to add a link here.
ReTweet
Stumble It!
Bookmark in del.icio.us
Support forum (1)
Blog
Latest changes
Groups
User ratings
Files
Advertise on this site
Site map
Statistics
Site tips
Privacy policy
Contact
