1: \begin{abstract}

2:     Stochastic first-order methods such as Stochastic Extragradient (\algname{SEG}) or Stochastic Gradient Descent-Ascent (\algname{SGDA}) for solving smooth minimax problems and, more generally, variational inequality problems \eqref{eq:main_problem} have been gaining a lot of attention in recent years due to the growing popularity of adversarial formulations in machine learning. However, while high-probability convergence bounds are known to reflect the actual behavior of stochastic methods more accurately, most convergence results are provided in expectation. Moreover, the only known high-probability complexity results have been derived under restrictive sub-Gaussian (light-tailed) noise and bounded domain assumption~\citep{juditsky2011solving}. In this work, we prove the first high-probability complexity results with logarithmic dependence on the confidence level for stochastic methods for solving monotone and structured non-monotone \ref{eq:main_problem}s with non-sub-Gaussian (heavy-tailed) noise and unbounded domains. In the monotone case, our results match the best-known ones in the light-tails case \citep{juditsky2011solving}, and are novel for structured non-monotone problems such as negative comonotone, quasi-strongly monotone, and/or star-cocoercive ones. We achieve these results by studying \algname{SEG} and \algname{SGDA} with clipping. In addition, we numerically validate that the gradient noise of many practical GAN formulations is heavy-tailed and show that clipping improves the performance of \algname{SEG}/\algname{SGDA}.

3: \end{abstract}

4: