A Non-Technical Introduction to Neural Networks
By Richard McKeon
Let’s build a network out of simple electronic components and train it by adjusting the connection weights. My goal is to give you an interesting and fun introduction to this fascinating topic in an easy to understand, nontechnical way. If you want to understand neural networks without using calculus, or differential equations, pick up my book: Neural Networks for Electronics Hobbyists!
There are no prerequisites. You don't need an engineering degree, and you don't even need to understand high school math in order to understand everything we are going to discuss. In this book you won't see a single line of computer code.
For this project we are going to take a hardware approach using simple electronic components. The project we are going to build isn't complicated, but it illustrates how back propagation can be used to adjust connection strengths or "weights" and train a network. We do this manually by adjusting potentiometers in the hidden and output layers.
Even if you are not an electronics hobbyist, I bet you will be able to build the network and understand how it works. There are plenty of detailed step-by-step instructions, pictures, and diagrams to guide you. There are no high voltages involved, and you won't burn the house down, so give it a shot! Who knows, this may turn out to be a fun introduction to the electronics hobby.
After a few passes of backpropagation training you will see the network gradually learn to perform the recognition task. We train it to perform various logic functions and then suggest other projects you may want to pursue.
Neural Networks are modeled after biological computers like the human brain. Instead of following a step-by-step set of instructions, a neural network consists of a bunch of "neurons" that act together in parallel - all at once - to produce an output. How cool is that?
Figure 1 is a picture of the completed project and Figure 2 is a Fritzing diagram showing the wiring.
Figure 1. The Completed Project
Figure 2. Fritzing Diagram of the Wiring
This blog post specifically references my recent release, Neural Networks for Electronics Hobbyists. This book is for anyone who wants to know a little more about neural networks. We start off with an interesting nontechnical introduction to neural networks, and then we construct an electronics project to give you some hands-on experience training a network using back propagation.
The first two chapters give an introduction to neural networks, and the rest of the book is devoted to building a network out of simple electronic components and training it by adjusting the connection weights. My goal is to give you an interesting and fun introduction to this fascinating topic in an easy to understand, nontechnical way. If you want to understand neural networks without using calculus, or differential equations, this is the book for you!
About the Author
Richard McKeon holds degrees in both Mathematics and Electrical Engineering. He worked as an engineer for several years designing microprocessor based products and installing communication networks. He now lives in beautiful Prescott, Arizona. Since retiring he has been spending time pursuing his passion for writing, playing music and teaching. Rick is currently producing a series of books on music, nature and science.
Some of his other interests include hiking, treasure hunting, recreational mathematics, photography and experimenting with Microcontrollers. To learn more about this author visit his website at rickmckeon.com.
Learn more about in Richard McKeon's book, Neural Networks for Electronics Hobbyists: A Non-Technical Project-Based Introduction, now available in both digital and print formats.