Neural Network construction and research.

• A neural network in this case is a program that tries to simulate the neurons of the brain. By training it with known problems the network tries to find a solution.

• Most NNs have some sort of "training" rule whereby the weights of connections are adjusted on the basis of data. In other words, NNs "learn" from examples (as children learn to recognize letters in the alphabet from examples) and exhibit some capability for generalization beyond the training data.

• Here is the first neural network we have built so far.
• We have made major improvments to the basic network design.
• The network can now learn the logic instructions of OR, XOR, NOT-XOR, AND and more Without getting stuck like the original network desgin did.
• This is achieved using only three neurons!.( 2 x input, 2 x hiidden, 1 x output = 1 x Perceptron )
• You can download this working demo here. ( just unzip to a folder and click the exe file to run )

Full credits to James Lewis for the base code to build from.

Multi layer dynamic neural network working!!! Dec 24th 2001

• Multi layer Neural Network with visual output.

• Random training of multiple logic gate formulas solved in real time.

• 1 of 32 to 5 bits decoder successfully trained. A test to see if the net was working ok

• Download the 1 of 32 to 5 bits decoder demo and watch it learn. It will start over when the problem has been solved.

The color changes from red to green as the error decreases.              Next Page

If you have any questions email us Kiwi@Pitstock.com