Forecasting Treatment Responses Over Time Using Recurrent Marginal Structural Networks

by Bryan Lim, Ahmed Alaa and Mihaela van der Schaar, NeurIPS 2018

Usage instructions

To generate a new simulation, run hyperparameter optimisation and tests for the RMSN use :

bash test_rmsn.sh
This produces the raw MSEs into the "results" folder, and saves calibrated models into the "models" folder
Note that for the paper, results are imported in normalised root mean-squared error terms -> i.e. sqrt(mse) / 1150

Name		Name	Last commit message	Last commit date
Latest commit History 13 Commits
libs		libs
models		models
results		results
simulation		simulation
README.md		README.md
configs.py		configs.py
core_routines.py		core_routines.py
script_decoder_fit.py		script_decoder_fit.py
script_decoder_test.py		script_decoder_test.py
script_propensity_generation.py		script_propensity_generation.py
script_rnn_fit.py		script_rnn_fit.py
script_rnn_test.py		script_rnn_test.py
test_rmsn.bat		test_rmsn.bat
test_rmsn.sh		test_rmsn.sh