知识卡片
监督目标不一定全部由人工提供
监督训练使用输入及目标,目标可由人工标注、系统记录或数据自身构造。自监督不是没有标签信号,而是从数据构造预测任务;检索、情感等任务也可用不同训练方式,不能仅凭任务名固定监督类别。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- · 查看原书依据
There are many ways to classify machine learning problems; one of them is by looking at whether you have labeled data—examples of outputs of your model in addition to the inputs. In supervised learning, the model receives sample inputs with outputs that were generated or vetted by humans. These outputs are often called ground truth, or labels. An example of a label is the toxicity
参考来源
- 位置:《Natural Language Processing in Action, Second Edition》附录D《附录D:机器学习工具与技术》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/appendix-d.html)
- 结论依据:摘录中关于监督目标不一定全部由人工提供的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:There are many ways to classify machine learning problems; one of them is by looking at whether you have labeled data—examples of outputs of your model in addition to the inputs. In supervised learning, the model receives sample inputs with outputs that were generated or vetted by humans. These outputs are often called ground truth, or labels. An example of a label is the toxicity