Top 5: Things to know about machine learning

There are several different types of machine learning, and it pays to know a little about each. Here's a quick guide.

Top 5: Things to know about machine learning

Machine learning is the ability for a program to learn something without having to be programmed to learn that specific thing.

One of the most famous examples is a program that learned to recognize cats by being fed cat pictures. But there are different types of machine learning and it pays to know a little about each.

But instead of teaching you by feeding you pictures of cats, we'll just tell you five types of machine learning.

1. Supervised learning

This is broad category with a few subtypes. Essentially it means we train the algorithm on some correct examples. So in our cat example, we would show the algorithm some cat pictures until it gets the idea and can start recognizing cats in other pictures.

SEE: IT leader's guide to deep learning (Tech Pro Research)

2. Semi-supervised learning

This is a subtype where the algorithm is trained on both labeled and unlabeled data. In other words there are a bunch of pictures labeled as cats but there are also a bunch of unlabeled pictures which may or may not have cats. It uses the labeled one to help figure out the unlabeled ones until again it can start recognizing cats in any picture.

SEE: Managing AI and ML in the enterprise (ZDNet special report) | Download the report as a PDF (TechRepublic)

3. Reinforcement learning

This one only gives training data in response to actions. It's really good for things like driving or games. Proper actions are reinforced, improper ones lead to failure like losing the game.

4. Active learning

This is one where the algorithm can get labels for a limited amount of data, so it's best to make guesses when it thinks it will be right.

SEE: How Sephora is leveraging AR and AI to transform retail and help customers buy cosmetics (PDF download) (TechRepublic cover story)

5. Unsupervised learning

This is where no labeled data is involved at all and no feedback is delivered, so no reinforced learning. An example might be what's called General Adversarial Networks, where two neural networks compete and success in competition drives the learning. Its especially promising in security for detecting advanced or previously unknown types of attacks.

Well now you have learned how the machines learn and maybe this will help keep you smarter than the machines!

For more about machine learning, check out these articles from TechRepublic and our sister site, ZDNet: