Next: Bibliography
Up: Artificial Neural Networks Simulation,
Previous: MATLAB Reference
- The limitation of number of inputs in visualization tools is caused by the visual
perception of the human being. For networks with more number of inputs, transformations
can be applied to reduce the number of dimensions to some principal dimensions
(3 or 2 dimensions), by using principal component analysis and Fischer's linear
discriminant [Agogino99].
- The parameters about the architecture of the network such as number of input,
output and patterns can be obtained by analyzing the corresponding data set
files, rather than required to be defined in the problem specification file.
- Adding ability to process multi-layer network architectures, as well as single
layer (no hidden layer) networks.
Cengiz Gunay
2000-06-25