1: \begin{abstract}

2: This paper analyzes the training process of GANs via stochastic differential equations (SDEs). It first establishes SDE approximations for the training of GANs under stochastic gradient algorithms, with precise error bound analysis. It then demonstrates the convergence of GANs training via invariant measures of SDEs under proper conditions. This work builds theoretical foundation for GANs training and provides analytical tools to study its evolution and stability. 

3: \end{abstract}

4: