This repository contains a neural network built in Java for the course of OhjelmistoTekniikan Menetelmä (OTM) at the University of Helsinki. The network works on poly-dimensional input data however it comes with a ready built method to train and validate it on one-dimensional inpu data of a cosine curve.
The network uses the activation function of the sigmoid curve and the optimizer of gradient descent. It performs well on one-dimensional input data however the performance declines quickly as the dimensions increase. The current performance can be seen here.
The source code is located in the folder src. Instructions how to employ the network and basic knowledge of the network and output is located in the folder documentation.
To compile and run the code, in the root directory (current) perform the following command
mvn compile exec:java -Dexec.mainClass=UI.Main
The test code is located in the folder test For testing the program, the following command can be performed in the root directory
mvn test