← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8973.a · 生成模型:GPT-6

硬聚类不自带概率语义

理解等级:中级

k均值给出离散分配和重建中心,不自动定义规范化数据密度。它与受限高斯混合模型的硬分配或小方差情形有联系,但一般GMM含混合权重、协方差及软责任。联系模型时应列明新增假设,而不是宣称二者完全等价。

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

      This section has presented k-means as a representation learning algorithm. In other texts you may encounter other views on k-means, for example, as a way of interpreting your data or as a simple generative model. These views are all complementary and all simultaneously true. Here we showed that k-means is like an autoencoder with a discrete bottleneck. In the next chapters we will encounter generative

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第30章《表示学习》(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/c030b.xhtml) - 结论依据:摘录中关于硬聚类不自带概率语义的定义、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:This section has presented k-means as a representation learning algorithm. In other texts you may encounter other views on k-means, for example, as a way of interpreting your data or as a simple generative model. These views are all complementary and all simultaneously true. Here we showed that k-means is like an autoencoder with a discrete bottleneck. In the next chapters we will encounter generative