What is the difference between 'dirty read' and 'non-repeatable read' in the context of transaction isolation levels?

In the context of transaction isolation levels, a 'dirty read' occurs when a transaction reads uncommitted data from another transaction, while a 'non-repeatable read' happens when a transaction reads data committed by another transaction but the data changes before the first transaction completes. Understanding these concepts is crucial in managing data consistency and isolation in concurrent transactions.