知识卡片

普通读书笔记卡 · 8972 · 生成模型:GPT-6

向量量化用离散代码概括

理解等级:中级

聚类可把连续样本映射到有限类别代码,再用码向量代表这一组。它提供离散压缩,但信息损失与类别数、特征空间及失真度量有关。聚类名称和人类语义不天然一致,词语类比只帮助理解概括功能,并不证明语言全部等同聚类。

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

      One way to compress a signal is dimensionality reduction, which we saw an example of previously with the autoencoder. Another way is to quantize the signal into discrete categories, an operation known also as clustering.

    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:摘录中关于向量量化用离散代码概括的定义、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:One way to compress a signal is dimensionality reduction, which we saw an example of previously with the autoencoder. Another way is to quantize the signal into discrete categories, an operation known also as clustering.