Assignment 3
Boosting
Out: 10/08/2026.
In this assignment, we experiment with boosting as a method for improving the prediction accuracy of learning algorithms.
- Use scikit-learn but design your own implementation of the
AdaBoost algorithm.
As a bonus, compare this with the implementation of AdaBoost in scikit-learn.
- Test the boosting algorithm on the UCI ML
digit database.
Convert the multiclass dataset to a two-class dataset (for example: even versus odd, prime versus non-prime, etc).
- Choose at least two "weak" learners for your experiments.
For one of the weak learners, use the
decision tree classifier
which is supported by scikit-learn.
Then, choose another weak classifier of your choice (within scikit-learn)
and compare this with the decision tree weak classifier.
- Provide a plot showing how the training error of the AdaBoost classifier
changes during the boosting process.
Determine whether boosting works effectively and which weak classifier is better.
What to submit
Submit via email your source file(s) and the plot (as PDF). You may work in groups of two.
CS549
Repeat the above using a different real-world dataset (other than the digit dataset) of your choice.