Supervised Learning: Explained, Briefly
“Machine intelligence is the last invention that humanity will ever need to make.” — Nick Bostrom
If you use Google Photos, I’m sure you’ve seen the image above. If you don’t, you might’ve seen something similar on your smartphone. But have you ever wondered how this happens?
That’s where machine learning comes in, more specifically, supervised learning.
As the name states, supervised learning refers to machine learning algorithms that learn under a supervisor. The training of the algorithm works by guiding it through labelled data. For instance, images of cats are labelled “cats.” But how does it exactly work?
- e start by inputting millions of images of cats and animals that are not cats.
- Then, the agent would begin processing data and try to find similar patterns.
- Now, the agent will start to classify “cats” and “not cats”
The classification would not be 100% accurate, which is where supervision is helpful. We would provide feedback on how correctly the machine sorts out the images. With this feedback, the machine will start to form better relations and patterns between the images and the feedback. Over time, the machine would be able to practice its classification almost perfectly.
Not Fully There Yet…
Like everything, even supervised learning isn’t flawless. There are two common problems that may restrict us from using this type of algorithm.
→ Overfitting takes place when the algorithm perfects itself so much that it’s unable to classify new data. For example, it looks for specific details where it can mistake something like a tiger cat as “not cats.”
→ Unlabelled data surrounds the world. It’s surreal to think of labelling each and every piece of data in order to train machine learning algorithms to a level of perfection.
So can you be the one to fix these problems?
