Here are some of the assignments implemented for my Neural
Networks class. Most of them are available for download. All other examples were omitted because
of their limited usefulness.
1. Perceptron. This program can run one or four perceptrons
in different modes- step-by-step, animation or run. A new linearly separable task
is set on each run automatically.
Download it (68K)
2. Madaline. Classical XOR problem. You
can go in a step-by-step mode to observe
solution process or just run it .
Download it (67K).
3. Kohonen Net (SAM). This program simulates a Kohonen net covering
one or four regions. Parameters of the network can be changed dynamically during execution
so that behavior of the network can be observed.
4. Backpropagation. An example of using
backpropagation for style - independentcharacter recognition. Very simple set of features extracted from a character wasused, providing sufficiently good results (at least to get a grade for it ;-))
Download it (123K).
5. Application of neural networks in something
'real'. If you wonder how a neural net can be applied to a real - world problem you
can find some answers on my ALPS (Automated
Lumber Processing System) page. A neural network solution is used to recognize different
damages on wood.