rguo12 / awesome-causality-algorithms Goto Github PK

treatment and outcome should be causal inference. There are many more non-iid related algorithms in causal discovery, such as those mentioned in this post https://sadhik9.people.uic.edu/shishir_wcp20.pdf, and I wonder if you'll be adding to this section subsequently.

Hi,

Would the following frameworks work for this repo?

Timeseries non-i.d.d. data

• TCDF: Temporal Causal Discovery Framework (https://github.com/M-Nauta/TCDF)

Structured Learning i.d.d. data

• DAG-GNN (https://github.com/fishmoon1234/DAG-GNN)
• DAGs with NO TEARS (https://github.com/xunzheng/notears)

Thanks!

Hey :)

I wanted to propose adding the X-learner meta-learner framework for heterogeneous treatment effects.

Paper: https://www.pnas.org/content/pnas/116/10/4156.full.pdf
Code: https://github.com/xnie/rlearner/blob/master/R/xlearner.R

I guess this belongs to the i.i.d data section.

Thanks for the awesome collection and the comprehensive review paper, great work!

Hi! I'm not sure if I'm supposed to fork the repo or just leave a comment, but do you mind adding this paper to the NLP section:

https://www.aclweb.org/anthology/2020.acl-main.474.pdf (ACL2020)

Thanks!

Shamelessly recommend my own preprint:
https://arxiv.org/abs/2101.06662
Some highlights: https://twitter.com/PengzhouW/status/1362225064309972993
In particular, it shows SOTA performance of balancing and deconfounding on IHDP and Pokec.

I do not know if preprints are allowed by the policy of your list. Simply ignore this if not.

I find CausalML is published with 2 pages in 2020.

@article{chen2020causalml,
  title={Causalml: Python package for causal machine learning},
  author={Chen, Huigang and Harinen, Totte and Lee, Jeong-Yoon and Yung, Mike and Zhao, Zhenyu},
  journal={arXiv preprint arXiv:2002.11631},
  year={2020}
}

I wanted to modestly propose expanding "Learning Time Varying/Dependent Causal Effects" section with our pre-print "Causal Transformer for Estimating Counterfactual Outcomes".

The code (https://github.com/Valentyn1997/CausalTransformer) also contains the PyTorch-Lightning implementation of other baselines over time, e.g., CRN, RMSN or MSMs.

I can create a corresponding pull request if you don't mind.

Best regards,
Valentyn

Thank you for this great list!

Currently there is only one reference under "Distinguishing Cause from Effect (Bivariate)".
Below are some references published in recent 3 years (in no particular order).
I try to make a high-variance-low-bias list, keeping an eye on the diversity, without bias to a particular kind of methods.
Hope these help.

https://dl.acm.org/doi/abs/10.1145/3292500.3330854 (KDD19)
http://proceedings.mlr.press/v84/bloebaum18a.html (AISTATS18)
http://proceedings.mlr.press/v119/tagasovska20a.html (ICML20)
https://link.springer.com/chapter/10.1007/978-3-319-98131-4_3
https://proceedings.neurips.cc/paper/2018/hash/73fed7fd472e502d8908794430511f4d-Abstract.html
https://proceedings.neurips.cc/paper/2020/hash/a979ca2444b34449a2c80b012749e9cd-Abstract.html
https://arxiv.org/abs/2007.02809 (AAAI21)
http://proceedings.mlr.press/v108/wu20b.html (AISTATS20)
This last is my own work, and to the best of my knowledge, it still keeps the best result on benchmark dataset.
I will publish the code in the recent future.

Cheers,
Pengzhou (Abel)

Hi,
I find one comprehensive survey paper about Potential Outcome Framework. Link is attached here.
https://arxiv.org/abs/2002.02770