It’s a computer system that mimics our brains. It’s software made up of connected nodes of information and every time a signal travels from one node to another it makes a connection that leads to an understanding of what’s happening in the situation.
Think of it this way: A neural network is a mall food court. Each restaurant option is a node. And you’re a waiter or waitress bringing food to people sitting at tables. They are also nodes. So you’re going back and forth between people and the restaurants bussing food. And you do that enough that you start to learn things about certain situations—say that 16-year-old males typically order the biggest, greasiest burger from the burger stand. This happens all the time. So the next time a 16-year-old male sits down at a table you grab the burger stand menu before you even head to his table.
Neural networks learn in the same way. They see connections between nodes happening more often than not and begin to make predictions.
“Over time, if the neural network uses certain connections more often, then they get stronger and stronger,” Dr. Wilfried Achenbach, Daimler Trucks North America’s senior vice president of engineering and technology explains, “but you only have a deterministic answer on a neural network concerning objects that the network is already trained to detect.
“Let’s say you train it, for instance, to detect bicycles, and then compare bicycles to cars and to trucks. If there’s a street scooter on the road, then it doesn’t know how to handle a street scooter. It had to be trained for the street scooter.”
