A Neural Network Primer

A Neural Network Primer

This paper gives the basics of the subject of neural networks. It is recommended for those new to the subject of neural networks and only assumes that the reader has a knowledge of basic algebra.

Please feel free to distribute this paper as you wish but please distribute the entire paper without deletions and limit your corrections and additions to the space below the bottom line after the last paragraph of the original text. If you wish for your correction or modification to be changed in the body of the original text, please send your recommendation to one of my computer addresses above.

A function is a mysterious black box with inputs and an output.

----------

X --->| |

| Function |---> Z

Y---->| |

----------

The above function has 2 inputs, X and Y, and one output, Z.

Mysterious black boxes that take in inputs and produce an output are called "functions". Sometimes we don't know what goes on inside the function to produce the output from the inputs so that's why we call them "black boxes".

X Y | Z

-----------

0 0 | 0

0 1 | 1

1 0 | 1

1 1 | 1

The above table maps out a possible function. It is called the "Or" function. Note that Z is zero when X and Y are both zero and Z is one whenever X "or" Y is a one. Mapping out a function in a table as is shown above is known as creating a "truth-table".

X Y | Z

-----------

0 0 | 0

0 1 | 0

1 0 | 0

1 1 | 1

The function mapped out in the truth-table above is known as the "And" function. Note that Z equals 1 only when X "and" Y both equal 1.

Note that the numerical values we have used so far for X, Y, and Z have been limited to only "0" and "1". When you only use two values like this you are using a "binary" or "boolean" number system. "Binary" and "boolean" basically mean "two values".