← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 6563.b · 生成模型:GPT-6

条件样条的单调轴要说清

理解等级:高级

条件样条以x作为context产生变换参数,保证的是固定x时对被变换噪声n的相应可逆单调性质。context网络通常不限制x方向,所以它不自动保证固定n时f(x,n)随x递增,更不保证饱和或边际递减。原书将两种单调性混用了。

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

      print(conditional_spline(input_dim=1, context_dim=1)) #1 #1 A neural spline transform is a type of invertible PyTorch neural network module. We get a three-layer neural network with ReLU activation functions: ConditionalSpline( (nn): DenseNN( (layers): ModuleList( (0): Linear(in_features=1, out_features=10, bias=True) (1): Linear(in_features=10, out_features=10, bias=True) (2): Linear(in_features=10, out_features=31, bias=True) ) (f): ReLU() ) ) Normalizing flows solve our problem of not having a likelihood value for y = f(x, ny).

    参考来源

    - 位置:《Causal AI》第06章《结构因果模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-6.xhtml) - 结论依据:摘录中关于条件样条的单调轴要说清的定义、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:print(conditional_spline(input_dim=1, context_dim=1)) #1 #1 A neural spline transform is a type of invertible PyTorch neural network module. We get a three-layer neural network with ReLU activation functions: ConditionalSpline( (nn): DenseNN( (layers): ModuleList( (0): Linear(in_features=1, out_features=10, bias=True) (1): Linear(in_features=10, out_features=10, bias=True) (2): Linear(in_features=10, out_features=31, bias=True) ) (f): ReLU() ) ) Normalizing flows solve our problem of not having a likelihood value for y = f(x, ny). - 外部校正依据:[Pyro ConditionalSpline官方文档](https://docs.pyro.ai/en/stable/distributions.html#pyro.distributions.transforms.ConditionalSpline)。文档分别定义被变换输入与生成样条参数的context网络;本卡据此指出原例未限制context方向的导数。