Approximate Allocation Matching for Structural Causal Bandits with Unobserved Confounders
Authors: Lai Wei, Muhammad Qasim Elahi, Mahsa Ghasemi, Murat Kocaoglu
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We complement the theoretical results by evaluating the proposed approach in a variety of experimental settings featuring different causal structures. We show that our algorithm outperforms the existing baselines in terms of empirical regret. We compare the empirical performance of SCM-AAM with existing algorithms. |
| Researcher Affiliation | Academia | Lai Wei Life Sciences Institute University of Michigan weilatim@gmail.com Muhammad Qasim Elahi Electrical and Computer Engineering Purdue University elahi0@purdue.edu Mahsa Ghasemi Electrical and Computer Engineering Purdue University mahsa@purdue.edu Murat Kocaoglu Electrical and Computer Engineering Purdue University mkocaoglu@purdue.edu |
| Pseudocode | Yes | Algorithm 1: SCM-based Approximate Allocation Matching (SCM-AAM) |
| Open Source Code | Yes | The algorithm code is provided at https://github.com/Causal ML-Lab/SCM-AAM. |
| Open Datasets | No | The paper describes simulation setups (Task1, Task2, Task3) with structural equations for data generation, but does not use or provide access to a pre-existing public or open dataset. |
| Dataset Splits | No | The paper describes simulation parameters like |
| Hardware Specification | Yes | The simulations were conducted on a Windows desktop computer featuring a 12th generation Intel Core i7 processor operating at 3.1 GHz and 32 GB of RAM. No GPUs were utilized in the simulations. |
| Software Dependencies | No | The paper mentions providing algorithm code but does not list any specific software dependencies or their version numbers within the text. |
| Experiment Setup | Yes | We choose the input parameters of the SCM-AAM algorithm to be = 0.1 and = 1/|5I| in the simulations. We set the horizon to 800 for all three tasks and repeat every simulation 100 times. |