em

paper/sec_why_domainlab.tex

@@ -8,7 +8,7 @@ \section{Statement of need}\label{statement-of-need}
 state-of-the-art methods.
 
 \emph{DomainBed}~\cite{domainbed2022github} as an early published solution provided a common codebase for benchmarking domain generalization methods~\cite{gulrajani2020search}, however applying its algorithms to new use-cases requires extensive adaptation of its source code, including, for instance, that the neural network backbones are hard coded in the codebase itself and all components of an algorithm have to be initialized in the construction function, which is not suitable for complex algorithms that require flexibility and extensibility of its components like~\cite{mahajan2021domain, sun2021hierarchical}.
-A more recent concurrent work, *Dassl*~\cite{dassl2022github}, provides a codebase for some domain adaptation and domain generalization methods with semi-supervised learning~\cite{zhou2021domain}.
+A more recent concurrent work, \emph{Dassl}~\cite{dassl2022github}, provides a codebase for some domain adaptation and domain generalization methods with semi-supervised learning~\cite{zhou2021domain}.
 Its design is more modular than DomainBed. However, the code base does
 not appear to be well-tested for long-term maintenance and the
 documentation is very limited.