知识卡片
预训练是知识获取阶段
预训练既可能是后来被复用的普通监督训练,也可以是专为迁移设计的表征学习。将预训练和适配分开,能说明何处付出数据成本、何处使用知识;它们不是由损失名称唯一决定的两个固定算法。
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
- · 迁移学习与适配
原书出处 · 1 处
- · 查看原书依据
The pretraining phase could just be regular training, which we only call pretraining in retrospect, once we decide to start adapting to a new problem. Or we can specifically design pretraining algorithms that are only meant as a warm up for the real event. Representation learning algorithms (chapter 30) are of this latter variety.
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第37章《迁移学习与适配》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c037.xhtml)
- 结论依据:摘录中关于预训练是知识获取阶段的定义、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:The pretraining phase could just be regular training, which we only call pretraining in retrospect, once we decide to start adapting to a new problem. Or we can specifically design pretraining algorithms that are only meant as a warm up for the real event. Representation learning algorithms (chapter 30) are of this latter variety.