Generic application set-up

Building an application for some experiments may take some time. Besides understanding the library and how to call all the functions needed, programming has to be done to implement the specific experiments you want to perform. To help the user to get started, a generic application set-up is written, which can serve as a starting point to build an SPRANNLIB application. Some trivial things like command line handling and loading and saving of data sets are already taken care of. The user only has to program his/her experiments and print out the desired statistics.

The library is fully documented with on-line manual pages. If however the usage is not completely clear, for most library functions an example usage can be found. These example applications are part of the library and can also serve very well as an initial set-up for a new program. The following example programs are implemented:

• Simple backpropagation (bpexam.c)
• Two Kohonen examples (kohdemo1.c and kohdemo2.c)
• Conjugate gradient descent method (netcdes.c)
• Levenberg Marquardt method (netmarq.c)
• Simple perceptron (pctron.c)
• K-nearest neighbour classification (knndemo.c)
• Maximum likelihood estimation (mlihood.c)