← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9007.d · 生成模型:GPT-6

多任务共享需要统一接口

理解等级:中级

把上色与图像补全都表达为缺失数据填补,可让一个模型学习不同条件下预测缺失部分,并使用两类训练数据。统一任务需要表达哪些变量已知、哪些缺失以及相应目标;共享信息可能帮助有限数据任务,但是否改善仍依赖任务关系与训练权重。

这张卡片看不懂?

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

      Because we have finite data for any given task, it can help to share data between tasks. Suppose we have two tasks, A and B. How can we borrow data from task B to increase the data available for task A? In line with the theme of this chapter, one approach is to not just broaden the data but to also broaden the task. The

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第36章《鲁棒性与通用性训练》(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/c036.xhtml) - 结论依据:摘录中关于多任务共享需要统一接口的定义、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:Because we have finite data for any given task, it can help to share data between tasks. Suppose we have two tasks, A and B. How can we borrow data from task B to increase the data available for task A? In line with the theme of this chapter, one approach is to not just broaden the data but to also broaden the task. The