Explain the difference between 'ovo' and 'ovr' in multi-class classification.

'ovo' (One-vs-One) creates a binary classifier for every possible pair of classes. 'ovr' (One-vs-Rest) creates a binary classifier for each class against all other classes. 'ovo' is generally more computationally expensive but can be more accurate for smaller datasets. 'ovr' is more efficient for larger datasets and often performs well. Choose the method based on your dataset size and desired performance.