j-mark-hou / npiv Goto Github PK

Implementation of the core idea behind the DeepIV paper, but with tree boosting and quantile regression instead of deep learning and mixtures of gaussians.

Just browsing:

• for more in depth discussion, see this blog post http://jmarkhou.com/npiv/
• look at example/example.ipynb

Run via docker: docker-compose up

How to build: docker-compose build

Development:

• dockerfile just installs the package via pip with the -e option, and docker-compose then mounts this directory to that install location, so that any changes made are directly reflected. Thus, just run the container, and make whatever changes as you like.
• to run tests as you develop, do docker exec 270033469fd0 pytest -s, excep with whatever the hash of the container is (do docker container ls to see running containers)
• or even better, just use bash in the container: docker exec 270033469fd0 bash