RNN Arithmetic

This project is described in detail here.

Description

This is a multilayer Recurrent Neural Network (RNN) that can learn how to perform basic arithmetic (binary addition so far).

In other words, you can feed it two binary strings, X and Y, and a target string, T. In the case of binary addition, T = X+Y :

X: 1 0 1 0 1 1 1 1 1 0 1 0 1 1
Y: 0 1 0 0 1 1 1 0 0 1 1 1 1 0 
T: 1 1 1 1 1 1 1 0 0 0 1 0 0 1
                             ^ 
                             |
                        <== RNN works from the LSB to the MSB

At the start, the RNN knows absolutely nothing about binary addition. However, after seeing enough inputs & targets, it can figure out how to perform binary addition for itself!

This RNN uses back propagation through time (BPTT) for training, and has the following structure:

Usage

1. Download the jar file.

2. In the directory containing the jar file, run:

   java -jar RNN-Arithmetic.jar
   

3. Enter help to view the list of available commands.