Skip to content

hudsonchen/F2BMLD

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

42 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Fully-First order Bilevel Mean-field Langevin Dynamics (F$^2$BMLD)

This repository contains the implementation of the code for the paper "Towards a Unified Analysis of Neural Networks in Nonparametric Instrumental Variable Regression: Optimization and Generalization".

Installation

To install the required packages, run the following command:

pip install -r requirements.txt

Reproducing Results

Offline Policy Evaluation

To reproduce the results for the experiment (Figure 1), run the following command:

python main/run_f2bmld.py --policy_noise_level 0.0 --noise_level 0.2 --lagrange_reg 0.3 --seed 0 --max_steps 50_000 --batch_size 32

You can vary the Lagrangian multiplier lambda by altering the argument '--lagrange_reg 0.3', vary the environment noise level by altering the argument '--noise_level 0.2'.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published