Given a dataset containing the features (weigth, country, year, etc.) of several cars, compare the architectures of different Multi-Layer Perceptron and Radial Basis Function networks built with MATLAB and choose the one which gives the smallest error.
Not all the features in the dataset are useful for the training of the MLPs and for this reason a subset of them have been choosen by using a Genetic Algorithm able compare the correlation of different subset of features with the MPG.
After choosing the features, a set of possible architectures for RBF and MLP networks have been evaluated by training them using the choosen set of features.
The whole process (features selection + network evaluation) can be executed using the script main.m
.