Causal reasoning in the spring - practical HTE (Heterogeneous Treatment Effects) collection of papers github

Has always been a problem-solving machine learning hope is that 'what if', that is to guide decision-making:

• If I were to send the user a coupon user will stay Why?
• If the blood pressure in patients taking this drug will reduce it?
• What if the long-APP add this feature will increase the user's use?
• If the effective implementation of the monetary policy to boost the economy it?

The reason why this kind of problem is difficult to solve because the ground truth in reality is unobservable, a blood pressure drug has served to reduce but we do not know if he is not also reduce blood pressure without medication at the same time.

This time the students should do the analysis that we do experiments AB! We estimate that the overall difference is remarkable effective, is not significant invalid. But we can do is that all?

of course not! Because each individual is different! Overall invalid does not mean that the local population is not valid!

• If only 5% of users send sensitive coupons, we can only touch it up to these users? Or different users of the coupon sensitive threshold values ​​are different, how to attract more users by adjusting the threshold value of the coupon?
• If antihypertensive drugs is only effective for patients with specific symptoms, and how do we find these patients?
• New features of APP some users do not like, some users really like, I can compare the differences of these users to find directions to improve this new feature it?

Try the following from a different perspective to solve this problem, but the basic idea is the same: we can not observe each user's treatment effect, but we can find a group of similar users to estimate the effects of their experiments.

I'll be following the blog, from The second Recursive partitioning for heterogeneous causal CasualTree the effects began to comb the similarities and differences of the following methods.

The whole field is still developing, are just a few open source release soon, so this blog will be continuously updated. If you see a good article and project realization are also welcome in the comments below ~

Uplift Modelling

1. Nicholas J Radcliffe and Patrick D Surry. Real-world uplift modelling with significance based uplift trees. White Paper TR-2011-1, Stochastic Solutions, 2011.[文章链接]

2. Yan Zhao, Xiao Fang, and David Simchi-Levi. Uplift modeling with multiple treatments and general response types. Proceedings of the 2017 SIAM International Conference on Data Mining, SIAM, 2017. [文章链接] [Github链接]

Casual Tree

1. Athey, S., and Imbens, G. W. 2015. Machine learning methods for
   estimating heterogeneous causal effects. stat 1050(5) [文章链接]

2. Athey, S., and Imbens, G. 2016. Recursive partitioning for heterogeneous causal effects. Proceedings of the National Academy of
   Sciences. [文章链接] [Github链接]

3. C. Tran and E. Zheleva, “Learning triggers for heterogeneous treatment effects,” in Proceedings of the AAAI Conference on Artificial Intelligence, 2019 [文章链接] [Github链接]

Meta Learning

1. M. Oprescu, V. Syrgkanis and Z. S. Wu. Orthogonal Random Forest for Causal Inference. Proceedings of the 36th International Conference on Machine Learning (ICML), 2019 [文章链接] [GitHub链接]

2. Sören R. Künzel, Jasjeet S. Sekhon, Peter J. Bickel, and Bin Yu. Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the National Academy of Sciences, 2019. [文章链接] [GitHub链接]