Unsupervised versus Supervised Learning
We'll take a moment to understand the difference between Unsupervised AI Learning versus Supervised AI Learning. Supervised neural networks must be told of correct answers to a question at some point beforehand, so that this type of AI is "taught" (ahead of time) what an object is for example, just like a baby learns objects from flash cards. This approach of Supervised Learning is typically used for object recognition in Deep Learning AI, such as to tell the difference between a cat from a terrorist bomber. This is great if you have 10,000 images of cats that you can show to your Deep Learning AI system, but if you have 10,000 images of cats, you probably have other problems we won't go into here.
Unsupervised Learning on the other hand, involves exposing all types of information to your AI system and relying on it to learn something you haven’t programmed it to recognize, since an Unsupervised Learning system would be able to cluster data into logical patterns.
For object recognition, Unsupervised Learning groups certain related shapes together and assumes that they are similar. It is often also used in AI for game playing, such as chess, using experience from playing opponents to group successful techniques for certain circumstances, to use in the future against others in similar circumstances.
For more information, see TechWorld's Big Data Article:
So for Multilayer Perceptron (MLP) Deep Learning AI, we'll use Supervised Learning, since MLP was designed to work well with the ahead-of-time teaching approach of Supervised Learning. Get those flash cards ready...