What is Bias Variance Trade-Off in Machine Learning?

Are you confused with the term “Bias Variance Trade-Off“? If yes, then don’t worry. In this article, I am gonna share What is Bias Variance Trade-Off in Machine Learning in a super easy approach. So give your few minutes to this article and understand the concept of Bias Variance Trade-Off.

Now without any further ado, let’s get started-

What is Bias Variance Trade-Off in Machine Learning?

Before moving to Bias Variance Trade-Off, first, understand what is Bias and Variance in machine learning?

What is Bias and Variance in Machine learning?

Supervised machine learning algorithms need a supervisor to learn and this supervisor is nothing but the training data. The training data is used to train the supervised machine learning algorithms.

In supervised machine learning, there are two values. One is the Actual value and another one is the Predicted value. The actual value is the value that we already have and the Predicted value is the value that the model predicts.

What is Bias?

Bias is the gap between the actual value and the predicted value. You might have heard High Bias and Low Bias. So the next question is- What are High Bias and Low Bias?

High Bias means the predicted value is far from the actual value. Or you can say there is a huge gap between the predicted value and the actual value.

Whereas Low Bias means the predicted value is closer to the actual value or there is a small gap between the actual and predicted value.

So you understood Bias, low bias, and high bias. Now let’s move to the Variance.

What is Variance?

Variance means how much-predicted values are scattered in relation to each other. In Variance, there is a Low Variance and High Variance.

Low variance means all the predicted values are in a close group and very close to each other. Whereas in high variance all the predicted values are far from each other.

Now let’s understand Low Bias, High Bias, Low Variance, and High Variance with the help of this image-

Note- In this image, the smallest circle is the target value.

So let’s start with scenario 1 (Low Bias and Low Variance)– As I mentioned earlier that low bias means the difference between the actual and predicted value is less. And Low variance means the predicted values are in a close group. That’s why you can see in scenario 1 all predicted values are closer to the target value(small circle) and forms a close cluster.

In scenario 2( Low Bias and High Variance)– Due to low bias, all the predicted values are closer to the target value(small circle) but due to high variance, the values are scattered and far from each other.

Let’s move to scenario 3( High Bias and Low Variance)– Due to high bias, the predicted values are far from the target value(small circle) but due to low variance, they are in the closed group.

In the last scenario 4(High Bias and High Variance)- all the predicted values are far from the target value because of high bias and far from each other due to high variance.

I hope now you understood the whole concept of Bias and Variance in machine learning. So let’s move to the Bias-Variance Trade-Off in Machine Learning.
Click to continue reading...