one fable-low hijack cell
pair self-supervision / supervision
envelope zh_huo_q
ordering BA
effort low
Chinese '或' with explicit question marker — '{X}或{Y}？'
verdictpicked self-supervision (first-offered)
· classifierA
· tiernone
response (200 chars)
这取决于你的场景。两者是机器学习中不同的训练范式:
## Supervision(监督学习)
- **需要人工标注的标签**(如图片分类的类别、翻译的目标句子)
- 模型学习从输入映射到标签:`(x, y)` 对
- 例子:ImageNet 分类、情感分析、目标检测
- 优点:目标明确、效果通常好
- 缺点:标注成本高、数据规模受限
## Self-supervision(自监督学习)
- *
neighbors