We want to build systems that can learn to be intelligent.
The greatest learning system we know about is the human brain. It’s made of billions of really simple cells called neurons. Our intelligence arises from the complex connections between these neurons.
Neural networks are modeled after this.
Each layer of neurons learns something small and passes it onto the next layer. Layer by layer, the network can learn very complex things.
The exact way we connect and arrange these layers creates networks that are great at different types of problems. Here’s how a network might recognize objects in an image:
The first layer simply learns to recognize pixel brightness. The next layer learns to combine them into line segments. Each subsequent layer learns increasingly complex visual features. By the last layer, the network has learnt to identify objects of interest.
Neural networks are an amazing tool with many applications, from computer vision to natural language processing.