It’s mentioned all the time in the news and companies are constantly touting how enhanced their applications are with learning machines. But what does that really mean?
To put it simply, its teaching a machine how to do something. There’s more than one way to do this but most models follow the same method of “Practice makes perfect”.
For example, if I wanted to learn how to draw, I would start with the basics and do simple things. I would practice the motions and gestures that I wanted to get the desired effect. The more I practice, the better (in most cases) I would get. My skill would improve over time. However I could be learning the wrong way and actually be making my drawing ability worse.
Teaching machines is the same way. A machine is given a data set (ex. a series of images it wants to learn how to draw) and it runs a loop that practices its skill at finding patterns amongst the data set (ex. a flower petal is more curves than straight lines, while a sidewalk is mostly straight lines). The way that computers do this falls under two main categories (supervised and unsupervised).
Supervised machine learning is when a machine is given feedback based on how they are functioning, and makes amendments accordingly. These are usually found in
Unsupervised machine learning is when a machine solves a problem of looks for a pattern without any assistance. It determines what it should be looking for on its own. These are commonly used in cases such as spam filtering and searching for anomalies in data.
The uses and benefits of Machine Learning extend far beyond the examples that I have listed above. The applications range from healthcare to music composition to determining movies that a user might like. Machine learning is becoming a growing part of the future and is advancing with each passing day.