知识卡片
候选文本决定分类边界
增加或改写类别描述会改变嵌入原型及分类结果,因此词句、候选集合和提示模板属于任务配置。相对最相似项不代表绝对匹配,尤其当所有候选都不合适时;需要独立处理拒识、歧义与置信校准。
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
- · 视觉与语言
原书出处 · 1 处
- · 查看原书依据
Figure 51.6 gives more examples of creating custom binary classifiers in this way. The two classes are described by two text strings (“triangle” versus “cricle”; “purple” versus “teal”; and “arrowhead” versus “ball”). The embeddings of the text descriptions are the red and green vectors. The classifier simply checks where an image embedding is closer, in angular distance, to the red vector or the green vector.
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第51章《视觉与语言》(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/c051.xhtml)
- 结论依据:摘录中关于候选文本决定分类边界的定义、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:Figure 51.6 gives more examples of creating custom binary classifiers in this way. The two classes are described by two text strings (“triangle” versus “cricle”; “purple” versus “teal”; and “arrowhead” versus “ball”). The embeddings of the text descriptions are the red and green vectors. The classifier simply checks where an image embedding is closer, in angular distance, to the red vector or the green vector.