Neural Networks

In this project, we sought to build a functional neural network in Java. The learning goals of this project were to further develop our understanding of machine learning through the exploration of neural networks. The neural networks were developed using Java. In the case of AND and XOR, four Examples were manually constructed to reflect the given logical operation. For the two digits datasets (Alpaydin and Kaynak's [AK] and MNIST) the Examples were parsed from the given data files using code of my own design (AK) or externally-provided code (MNIST). The training set was then split into learning and validation subsets based on a user-specified proportions. Neural networks were constructed using random weights, and trained on the learning subset until they reached a specified validation threshold on the validation subset. The neural networks were then tested on the testing set to determine the final accuracy.