知识卡片

普通读书笔记卡 · 8935 · 生成模型:GPT-6

编码解码器不保证互逆

理解等级:中级

编码器把输入转换为表示,解码器从表示生成数据域输出;CNN常分别通过逐级下采样与上采样实现。低维表示可节省计算并形成信息瓶颈,但可能丢细节。两者结构方向相反不意味着函数互逆,分割解码器的目标本来也不是恢复原始照片。

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

      In chapter 23, we encountered image pyramids that include both an analysis pipeline, which converts an image into a multiscale representation of filter responses, and a synthesis pipeline, which reconstructs the image from the filter responses. Deep networks also may operate in either the analysis direction or the synthesis direction. In the context of deep networks, we call the analysis network an encoder and the

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第24章《卷积神经网络》(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/c024b.xhtml) - 结论依据:摘录中关于编码解码器不保证互逆的定义、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:In chapter 23, we encountered image pyramids that include both an analysis pipeline, which converts an image into a multiscale representation of filter responses, and a synthesis pipeline, which reconstructs the image from the filter responses. Deep networks also may operate in either the analysis direction or the synthesis direction. In the context of deep networks, we call the analysis network an encoder and the