What is the greedy approach and when can it be applied?

The greedy approach makes the locally optimal choice at each step, hoping to find the globally optimal solution. It works when a problem has the greedy choice property, meaning a locally optimal choice leads to a globally optimal solution, and optimal substructure. Unlike DP which considers all options, greedy commits to one choice and never reconsiders. Greedy is correct for interval scheduling, Huffman coding, Dijkstra's shortest path, and minimum spanning trees. It fails when local choices do not guarantee global optimality, like the zero-one knapsack.