This coursework requires the use of the software Neurosolution. Here are the reqirements
This assignment requires you to write a technical report on a classification dataset of your choice. You should first give a background literature review on the topic of the dataset, you should then provide a design of the neural network used including justifications of design choices. You should comment on how well the network performed and cite the performance data such as the classification accuracy, confusion matrixes and performance under noise. Finally you should give observations on the suitability of neural networks to solve the chosen problem.
1. Background literature review
Your background literature review should explain the topic of choice and explore the problem in detail. The literature review should explain the input values, the classifications and any other important information about the problem.
2. Neural Network design choices
In this section you should explain the various design choices available to you for training your neural network. Such choices might include: Network architecture, hidden layer depth, the number of processing elements per layer, the use of genetic optimisation on: input space/No. of processing elements/learning rate/momentum. Your design should be informed by research wherever possible (cite references to relevant work) and should be critical. For example, consider the costs and benefits of certain design choices. Be careful to define and explain all terms.
3. Neural network performance
In this section you should present performance data obtained by the network including the time taken to learn, the classification accuracy of the network, the confusion matrixes values and the performance of the network under noisy conditions.
In this section you should comment on the suitability of neural networks to solve the selected problem and any other important information.