One Causality¶
一个人,一篇论文,一个视频,一个slide, 一个会议,一个教程,一句话,说明因果推断。
You can suggest and comment here
小图灵测试:How can machines represent causal knowledge in a way that would enable them to access the necessary information swiftly, answer questions correctly, and do it with ease, as a human can?
Judea Pearl¶
Judea Pearl is credited for causal diagrams. 更多详细内容参考:
更多因果研究工作者参考:https://sites.google.com/view/causal-inference-zerotoall/people
Seven Tools¶
这是是一篇必须背诵的综述和启发性文章。该论文首先总结了当前AI面临的三个主要困难,指出教会机器因果推理能够解决这些困难。然后提出构建因果引擎的三级因果思维,指出当前机器学习算法都停留在第一个层面。最后综述了因果研究的七大方面内容。 - 论文地址:The seven tools of causal inference, with reflections on machine learning. - 论文解读 Pear 2019 seven tools of causal inference
ACM Mar. 2019 - The Seven Tools of Causal Inference
该视频基本上就是 Judea Pearl 亲自向大家介绍他的论文 The seven tools of causal inference, with reflections on machine learning.
[1]:
from IPython.display import YouTubeVideo
YouTubeVideo('CsMV5o3hotY', width=800, height=400)
[1]:
Video: The new science of cause inference¶
我们推荐一个因果理论的介绍性视频。该一个多小时的视频是研究因果理论必看的。
[2]:
from IPython.display import YouTubeVideo
YouTubeVideo('ZaPV1OSEpHw', width=600)
# A Introduction to Causal Inference
# Keynote: Judea Pearl - The New Science of Cause and Effect
[2]:
Slide¶
Bernhard Schölkopf is a director at the Max Planck Institute for Intelligent Systems in Tübingen, Germany, where he heads the Department of Empirical Inference.
Bernhard Scholkpf 关于 Causality for Machine Learning 的一个介绍
https://drive.google.com/file/d/1YjkCTX_ns5ba4DvdNtfRcSBoAEFPXKuz/view
Tutorial¶
Tutorial on Causal Inference and Counterfactual Reasoning
Amit Sharma (@amt_shrma), Emre Kiciman (@emrek)
ACM KDD 2018 International Conference on Knowledge Discovery and Data Mining, London, UK
http://causalinference.gitlab.io/kdd-tutorial
Code: http://github.com/Microsoft/dowhy
该教程微软因果推理小组,在 ACM KDD 2018 一个关于因果推断非常详细清楚的介绍。Pearl的因果图模型是通向AGI的一个重要组件,但是由于其理论尚未完善(例如当前结果大部分都是建立在有向无环图的基础上,Bernhard 指出有环的情况难以处理),相关的实践并不多。相反在传统的因果推断理论在相关理论已经被广泛使用,该教程关注传统领域的因果效应估计问题,既讲清楚了因果理论的基础概念,又讲了很多直观简单的例子,所以非常推荐。