知识卡片
ReLU组合可构造局部凸起
通过平移、缩放四个ReLU并以[1,−1,−1,1]加权,可形成带窄斜坡的梯形凸起,再加权组合多个凸起近似曲线。斜坡宽度趋零只在适当点和误差意义下接近阶跃,连续网络不能在跳点上无条件一致逼近不连续指标函数。
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
- · 神经网络
原书出处 · 1 处
- · 查看原书依据
Figure 12.13: A bump can be represented as a weighted sum of shifted and scaled relu functions. Putting everything together, we have a linear-relu-linear for each bump, followed by a linear layer for summing up all the bumps. The two linear layers in sequence can be collapsed to a single linear layer, and hence the full function can therefore be approximated, to arbitrary precision, by
参考来源
- 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第12章《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/c012.xhtml)
- 结论依据:摘录中关于ReLU组合可构造局部凸起的定义、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 12.13: A bump can be represented as a weighted sum of shifted and scaled relu functions. Putting everything together, we have a linear-relu-linear for each bump, followed by a linear layer for summing up all the bumps. The two linear layers in sequence can be collapsed to a single linear layer, and hence the full function can therefore be approximated, to arbitrary precision, by