From 980cfc0f6690c802a7cdba6aecd8ea0fd390684c Mon Sep 17 00:00:00 2001
From: kristinblesch <48204979+kristinblesch@users.noreply.github.com>
Date: Fri, 22 Sep 2023 12:40:27 +0200
Subject: [PATCH] Update paper.md

---
 paper/paper.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/paper/paper.md b/paper/paper.md
index 5c11373..076c0fb 100644
--- a/paper/paper.md
+++ b/paper/paper.md
@@ -34,7 +34,7 @@ affiliations:
 
 # Summary
 
-Generative modeling is a challenging task in machine learning that aims to synthesize new data which is similar to a set of given data. State of the art are computationally intense and tuning-heavy algorithms such as generative adversarial networks [@goodfellow2014;@xu2019], variational autoencoders [@kingma2014], normalizing flows [@rezende2015], diffiusion models [@ho2020] or transformers [@vaswani2017]. A much more lightweight procedure is to use an Adversarial Random Forest (ARF) [@watson2023].  ARFs achieve competitive performance in generative modeling in much faster runtime [@watson2023] and are especially useful for data that comes in a table format, i.e., tabular data. That is because ARFs are based on random forests [@breiman2001] that leverage the advantages that tree-based methods have over neural networks on tabular data (see [@grinsztajn2022]) for generative modeling. Further, as part of the procedure, ARFs give access to the estimated joint density, which is useful for several other fields of research, e.g., unsupervised machine learning. For the task of density estimation, ARFs have been demonstrated to yield remarkable results as well [@watson2023].  Hence, ARFs are a promising methodological contribution to the field of generative modeling and density estimation. To reach scholars in these fields that are predominantely based in python, and a broad audience more generally, a fast and userfriendly implementation of ARFs in python is highly desirable, which is provided by the software package `arfpy`. 
+Generative modeling is a challenging task in machine learning that aims to synthesize new data which is similar to a set of given data. State of the art are computationally intense and tuning-heavy algorithms such as generative adversarial networks [@goodfellow2014;@xu2019], variational autoencoders [@kingma2014], normalizing flows [@rezende2015], diffiusion models [@ho2020] or transformers [@vaswani2017]. A much more lightweight procedure is to use an Adversarial Random Forest (ARF) [@watson2023].  ARFs achieve competitive performance in generative modeling in much faster runtime [@watson2023] and are especially useful for data that comes in a table format, i.e., tabular data. That is because ARFs are based on random forests [@breiman2001] that leverage the advantages that tree-based methods have over neural networks on tabular data (see @grinsztajn2022) for generative modeling. Further, as part of the procedure, ARFs give access to the estimated joint density, which is useful for several other fields of research, e.g., unsupervised machine learning. For the task of density estimation, ARFs have been demonstrated to yield remarkable results as well [@watson2023].  Hence, ARFs are a promising methodological contribution to the field of generative modeling and density estimation. To reach scholars in these fields that are predominantely based in python, and a broad audience more generally, a fast and userfriendly implementation of ARFs in python is highly desirable, which is provided by the software package `arfpy`. 
 
 # Statement of need