Underfitting is when something learns too little and can't do its job well because it didn’t pay enough attention to the details.
Imagine you're trying to draw a dog, but all you see is a ball. You think the dog looks like a ball, so you draw a perfect circle. But if someone shows you a real dog, you might be confused, it’s not a ball! That's underfitting: your drawing didn’t learn enough about what a dog actually looks like.
Why It Happens
Sometimes, when we try to solve a problem, we don't use enough clues or examples. Like if you only saw one kind of dog before, maybe a poodle, and thought all dogs look the same. Then you'd draw a poodle every time, even if it's actually a bulldog.
What It Looks Like
It’s like trying to solve a puzzle with just two pieces when there are twenty. You might think you've got it figured out, but you're missing most of the picture. Your answer is simple, but not quite right, it doesn’t match what you’re supposed to learn.
Examples
- A student studies only one chapter for an exam with ten chapters, they can't answer most questions.
Ask a question
See also
- How Does Complete Guide to Cross Validation Work?
- But What Is Overfitting in Machine Learning?
- How Does Interpretable vs Explainable Machine Learning Work?
- How Does Machine Learning Fundamentals: Cross Validation Work?
- How Does Machine Learning Explained in 100 Seconds Work?