MathWiz86 / Neural-Network-Demo

A demo of a neural network with one input, one output, and one hidden layer. This was for my Fuzzy Logic and Neural Network class.

Neural Network Demo

A basic simulation of a Neural Network with one input, one output, and one hidden layer. This takes in a .csv file with inputs between 0 and 1, with their expected outputs. Then it outputs the actual outputs to the .csv file.

The target equation is:

f : [0, 1] => f(x) = sin(x)

The neural network's equation is a sigmoid:

     x
-----------
  _________
\/1 + (x*x)

Opening the Project

This project is compatible with Unity versions 2020.2 and upwards.

How to Use

• On the 'Main Controls' screen, click 'Run Network'.
• You can click on elements to the left, and then click 'Remove Element' to remove them.
• You can add elements by clicking 'Add Element'.
• You can automatically solve all expected outputs by clicking 'Solve Inputs'
• You can manually edit each element's input and expected output. It will save to file if 'Auto-Save' is on.
• When the graph is made, you can hover over a node to see its Input (X) vs Output (Y).

Notes

• By default, there is no data. Opening the executable will give the option to create basic data.
• Data is stored at '_Data\neural_data.csv'.
• You can change the Neural Network settings in 'Options'.
• Auto-Save is on by default. You can change this in 'Options'.