tianyuan / gitblog

My drafts using issues and GitHub Actions.
0 stars 0 forks source link

Book of Why和线性系统的碎碎念 #20

Open tianyuan opened 10 months ago

tianyuan commented 10 months ago
  1. 和家属聊book of why这本科普书,讨论什么是causal,我们一致认为比较简单的入口是自问:what is the counterfactual? 只有找对reference group (the counterfactual),才能找到causal谜团的入口 (这需要数学统计的训练,也需要domain knowledge和经验). 我觉得想明白counterfactual(或者至少知道这个concept)是衡量一个人是否有critical thinking的一个很好的measure。

  2. 关于DAGs, 我总觉得它和Systems Dynamics里一直使用的causal loop diagram其实很像。

  3. 听了Gilbert Strang上的一个podcast片段. 有人问他微积分和线性代数,哪个应该先学。老爷子觉得线代应该放在前面,"it is all about flat things. If everything is all flat, you can't go wrong". 即便是曲线的世界(微积分),其实我们也是通过linearity来理解的(求切线和slope都是基于linear系统的,曲线就是一堆piecewise linear steps,然后你再fold这些linear steps,各种优化也是转化成线性问题), 线性空间更简单,也更好理解,其实应该放在微积分的前面。这个观点很新颖,确实有道理~

  4. 顺着线性代数和线性系统,忍不住想讲个冷笑话,不知道几个人能get!There is a reason why people belief in "earth is flat" (the flat-earth theory), because it is a linear system. We learn the "curved" world through linear systems. If everything is all flat, you can't go wrong (quote: Gilbert). LOL, 我自己先笑为敬。