What is ensemble diversity and how is it achieved?
Updated May 16, 2026
Short answer
Ensemble diversity refers to differences among models' predictions, improving generalization when combined.
Deep explanation
Diversity is essential in ensemble systems because it ensures that models do not fail on the same inputs. Diversity can be achieved through different training datasets (bagging), different feature subsets (random subspace), different algorithms (heterogeneous ensembles), or different hyperparameters. The tradeoff is between accuracy of individual models and diversity of the ensemble. Optimal ensembles balance both to minimize generalization error.
Real-world example
Autonomous driving systems combine vision, radar, and lidar models for diverse perspectives.
Common mistakes
- Using identical models and expecting ensemble gains.
Follow-up questions
- Is diversity always good?
- How do you balance diversity and accuracy?