To view this video please enable JavaScript.
Course Description
Although we frequently think of machine-learning modelling as a process to build a single model of the data, there are many times where building multiple models on the same dataset can give us some advantages. Bagging, or bootstrap aggregation, and boosting are two processes to do this. When building a single model, the optimization is done for the group as a whole. That means that there will be some smaller cohort for whom the prediction might be significantly worse. And often, practitioners aim to prevent this discrepancy in predictive accuracy.
Bagging and boosting are an important part of a practitioner’s toolkit because they train multiple models in different ways and can be used to tackle some of the ethical challenges in building effective models in terms of fairness. These lessons will review the differences between bagging and boosting and discuss the ways they can be used to improve a model’s prediction in smaller subgroups.
What You'll Learn
- Determine when it is appropriate to use bagging and boosting
- Explain the advantages of using bagging and boosting
- Distinguish bagging (bootstrap aggregation) from boosting as approaches to building multiple models
- Apply bagging and boosting to improve a model's prediction in smaller subgroups
- Address fairness and ethical challenges in model building using multiple-model methods
- Consider practical implementation factors for bagging and boosting
Key Takeaways
- Building multiple models on the same dataset can offer advantages over building a single model.
- Bagging (bootstrap aggregation) and boosting are two processes for training multiple models in different ways.
- Optimizing a single model for the group as a whole can leave some smaller cohort with significantly worse predictions.
- Bagging and boosting can be used to tackle ethical challenges in model building related to fairness.
- These methods can help improve a model's prediction in smaller subgroups.
Frequently Asked Questions
What does this course cover?
It reviews the differences between bagging and boosting and discusses how they can be used to improve a model's prediction in smaller subgroups. Lessons cover Bagging, Boosting, and Implementation Considerations.
What will I be able to do after this course?
You will be able to determine when it is appropriate to use bagging and boosting and explain the advantages of using them.
What is the difference between bagging and boosting?
Bagging, or bootstrap aggregation, and boosting are two processes for building multiple models on the same dataset, training multiple models in different ways. The course reviews the differences between them.
How do bagging and boosting relate to fairness?
Because they train multiple models in different ways, bagging and boosting can be used to tackle some of the ethical challenges in building effective models in terms of fairness, including improving prediction for smaller subgroups.
What skills will I gain?
The course covers skills including Boosting, Gradient Boosting, Machine Learning, Machine Learning Algorithms, Machine Learning Methods, and Machine Learning Model Training.
