Could anyone explain what greedy learning is in neural networks?