Online Balanced Experimental Design

Authors: David Arbour, Drew Dimmery, Tung Mai, Anup Rao

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we provide simulation evidence on the efficacy of our proposed methods.
Researcher Affiliation Collaboration 1Adobe Research, San Jose, CA, USA 2Data Science @ University of Vienna, Vienna, AT.
Pseudocode Yes Algorithm 1, takes i = 1, . . . , n unit vectors in sequentially and assigns them to a treatment and control
Open Source Code No The paper does not provide an explicit statement or link for the open-source code of the methodology described.
Open Datasets No All data generating processes used in simulations are shown in Table A1. If not otherwise specified, the sample size is 1000 subjects, the number of groups is two and the marginal probability of treatment is 1/2.
Dataset Splits No The paper describes simulation studies with various data generating processes but does not specify explicit train/validation/test splits common in ML model training.
Hardware Specification Yes All timings performed on a ml.r5.2xlarge instance of Amazon Sage Maker.
Software Dependencies No The paper describes algorithms and simulations but does not explicitly list specific software dependencies with version numbers (e.g., 'Python 3.8, PyTorch 1.9').
Experiment Setup Yes If not otherwise specified, the sample size is 1000 subjects, the number of groups is two and the marginal probability of treatment is 1/2.