My thesis or final year project was titled “Genetic Algorithms and Neural Networks”.
It was an investigation into the suitability of genetic algorithms for the training of neural networks when compared with back-propagation.
I also investigated the potential of the cascade algorithm (moving slightly outside the original project specification) and compared and contrasted the performance and other pros and cons of the three algorithms in my report.
The genetic algorithm that was used was a simple implementation limited to single-point crossover and no complex functionality, however it was entirely self-programmed (in C#). The back-propagation and cascade functionality was implemented via a C++/CLI DLL interfacing with unmanaged C++ classes that used functionality from the C version of the FANN Library.
The mark I achieved for this project was 71%, at the University of Glamorgan, South Wales.
The write up is downloadable in PDF format in three pieces (to satisfy the three milestones of the project process).
Milestone one
Milestone two
Milestone three – (was supported by a 20 minute presentation)
Supporting Application – (Only compiled for 32 bit Windows)
