問題文
A team must classify photographs of finished parts as acceptable or defective. One proposal writes explicit rules about acceptable dimensions. Another trains a deep network on labeled photographs. The team wants to state what the deep learning approach does differently. Which statement states it?
選択肢
- It produces a set of readable rules at the end of training, which the quality team can review and adjust.
- It learns the features that matter directly from the labeled images instead of having them specified by an engineer.
- It guarantees a correct answer for every image the line produces, because the network has already seen every possible kind of defect during its training.
- It reaches a correct answer without needing any labeled examples, which is why it suits inspection work.