What are the most common performance metrics used to evaluate machine learning models, and how do they differ for classification versus regression tasks?

What are the most common performance metrics used to evaluate machine learning models, and how do they differ for classification versus regression tasks?