Supervised and Unsupervised Learning For Kids
In this blog, I am trying to answer an interview question that I was asked during a data science interview.
An Interview question
Here is a very famous machine learning interview question:
What is the difference between supervised learning and unsupervised learning?
You can give an answer like this:
In supervised machine learning algorithms, we use labeled data, for example, classification of X-ray images as pneumonia, or not pneumonia.
In unsupervised machine learning, we don’t need to have labeled data, for example, clustering of customers to segments.
A Different Interview Question
What if the interviewer changes the question a little bit? The new question is not very different than the previous one. But the answer is quite different.
What is the difference between supervised learning and unsupervised learning? Explain to me as though I were a five-year-old.
You may need to change this answer. Because a five-year-old kid doesn’t know what labeled data is.
Let's try to answer this question using layman terms.
Imagine that Jack and John are given fruits to separate.
Jack was given bananas, strawberries, and apples. He knows what each fruit is. He classifies them using his previous knowledge and he labels each fruit as apple, banana, and strawberry. He is less likely to make an error.
John was given pomegranates, lychees, rambutans, and mangosteens. He didn’t see these fruits before. He separates them into groups, he makes “fruit segmentations” out of the given fruits. Because he doesn’t know the names of the fruits, he can not label them. He just looks at their shapes, sizes, and colors, maybe textures and smells. He makes groups(clusters) out of given fruits. John is more likely to make a mistake because he doesn’t have information about the classes.
Actually, in the data science world, Jack is doing supervised machine learning, namely classification, and John is doing unsupervised machine learning, namely segmentation.
You can practice answering similar questions in Layman's terms.
Explain [a data science topic] to me as though I were a five-year-old.
You are going to be asked this kind of questions in the data science interviews.
Kemal Yilmaz, thank you for asking and answering this question.
Thanks for reading.