Projects for the Artificial Intelligence (IART) class of the Master in Informatics and Computer Engineering (MIEIC) at the Faculty of Engineering of the University of Porto (FEUP).
Made in colaboration with Afonso Ramos and Julieta Frade.
Completed in 21/05/2018.
The program must properly train an Artificial Neural Network using the Back-Propagation algorithm, based on a set of data available for this purpose. The data set must be carefully analyzed in order to verify the possible need for pre-processing. The model obtained must then be used in the classification of new cases.
This project encompasses the following procedures:
- Designing a multi-layered neural network: the input layer contains the attributes or variables of data identification (which?), the output layer contains the classification obtained and the intermediate layer(s) assists in the functioning of the neural network. Several network configurations (number of layers, number of cells in different layers, input variables, learning algorithm parameters) must be tested, and their results are analyzed and compared in order to define the best architecture.
- Implementation / application of the algorithm "Back-Propagation".
- Detailed measurement of results in training and test data.