What is diversity-accuracy tradeoff in ensemble learning?

The diversity-accuracy tradeoff balances individual model accuracy with differences between models.

Ensembles benefit from diversity, but highly inaccurate models reduce performance. The tradeoff lies in choosing models that are both reasonably accurate and sufficiently different. Techniques like bagging increase diversity through data sampling, while boosting increases accuracy but may reduce diversity. Optimal ensembles balance both to minimize generalization error.