Advanced AI Systems - Neural Network Coursework Deadline: 23 March 2017
Implement a multi-layer perceptron (MLP) trained using the error backpropagation algorithm. Once implemented the network should be trained on the data set provided and its perfoemance should be evaluated.
There are a number of stages to this process that will be evaluated:
- Implementation of the algorithm in an appropriate language
- Documentation and commenting of this implementation
- Appropriate data pre-processing of the supplied data set
- Appropriate training, configuration and weight adjustment of the ANN model
- Appropriate evaluation of the chosen ANN structure
- A report detailing the above process and discussing the evaluation of the model
For additional marks the ANN model should be contrasted with a simple data driven model such as a multiple linear regression model (for example, see LINEST in Excel). You should also try different ‘improvements’ to the standard backpropagation algorithm and report on these – for example, momentum, annealing, etc.