Toying Around With Neural Networks
Neural networks have been around in one form or another for millions of years.
Up until fairly recently (from 1940 onward) they were only implemented in wetware inside of various living organisms. In recent decades it has become possible to implement neural networks on hardware and software to great effect in areas like speech recognition and computer vision.
There are many types of neural networks. The feedforward neural network has been extensively studied and is often used to provide an introduction to the concepts underlying neural network technology. The diagram below shows how an equation from introductory algebra can be interpreted as a feedforward neural network.
The feedforward neural network has “units” (the circles) that perform mathematical operations on the sum of outputs from connected units multiplied by the “weight” (the boxes) on the connection. The unit at the very end of the processing stream is the output unit which…
View original post 2,988 more words