This code in this project was written to empirically verify the main hypothesis of my masters thesis.
The thesis itself can be found at http://www.ai.rug.nl/~mwiering/Thesis_Tim_Kuipers.pdf
A neural network framework has been built with limited parallelism: the operations on 3D and 4D matrices are parallelized over the number of cores in your CPU.
Several objective functions formulated within the spread maximization paradigm which have been compared with established machine learning techniques for neural networks. The framework might therefore have a more broad applicability range than my thesis alone.
This code is released under the terms of AGPL3.