Original Source Here

So, we are able to understand that bias results in underfitting and varaince results in overfitting. Our ultimate aim is to find the right balance between bias and variance based on the problem statement to let our model converge better. And, be away from overfitting and underfitting problem.

Time to go through the various example based on the above graph which will definitely help you to understand in depth.

## High bias and high variance

This is the situation when your given dataset have high bias and high variance.

There is no point in guessing how your model will generalize in such cases.

## Low Bias High Variance

Data set is low biased but varying a lot. You can see in the diagram all the datasets are inside the circle but spread all across. Again, your model will have complex and will not able to converge well with unseen data.

## High Bias Low Variance

This data set is more biased and we can see clearly all the data points are together but not near to the center point. In other terms, model will be have bias error and will result in underfitting.

## Low Bias Low Variance

This is the ideal dataset where bias and variance are low. And, we can see in visualization that data points are together and very close to the centre point.

In the given case, resulted model will be simple and should be able to converge/generalize better.

Other factors also able to make their presence feel. But these few terms are important to understand and could be treated as base of ML to move ahead.

At low bias (data points are near to target) and low variance (data points are near to each other) model performs well.

# Conclusion

We would have understood that correct bias and variance trade off can result in better model which can generalize better. Other than that, it could result in underfitting and overfitting. We understood the concept here and it will definitely help to understand the model performance much before it goes to the final stage i.e, Production. These basic concepts work and will help you to give better and well trained model.

I hope this understanding will help you to understand your model performance. It will give you more confidence about your model to perform better in actual environment. Apply this knowledge and keep experimenting to solve problem statements. **Happy Coding and Learning!**

AI/ML

Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot