知识卡片
自动编码同时训练编解码
自动编码器由编码器f和解码器g组成,重建为g(f(x))。低维瓶颈限制传递信息,也可用稀疏、噪声或其他约束替代。降维映射通常无法在全输入空间精确求逆,解码器主要学习在数据分布上近似恢复,而非天然等于f的数学逆。
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
- · 表示学习
原书出处 · 1 处
- · 查看原书依据
To be precise, an autoencoder F consists of two parts, an encoder f and a decoder g, with F = g ◦ f. The encoder, f : ℝN → ℝM, maps high-dimensional data x ∈ ℝN to a vector embedding z ∈ ℝM. Typically, the key property is that M < N, that is, we have performed dimensionality reduction (although autoencoders can also be constructed
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:摘录中关于自动编码同时训练编解码的定义、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:To be precise, an autoencoder F consists of two parts, an encoder f and a decoder g, with F = g ◦ f. The encoder, f : ℝN → ℝM, maps high-dimensional data x ∈ ℝN to a vector embedding z ∈ ℝM. Typically, the key property is that M < N, that is, we have performed dimensionality reduction (although autoencoders can also be constructed