Euler-type methods for Levy-driven McKean-Vlasov SDEs with super-linear coefficients: mean-square error analysis

Abstract: We develop and analyze a general class of Euler-type numerical schemes for Levy-driven McKean-Vlasov stochastic differential equations (SDEs), where the drift, diffusion and jump coefficients grow super-linearly in the state variable. These numerical schemes are derived by incorporating projections or nonlinear transformations into the classical Euler method, with the primary objective of establishing moment bounds for the numerical solutions. This class of schemes includes the tanh-Euler, tamed-Euler and sine-Euler schemes as special cases. In contrast to existing approaches that rely on a coercivity condition (e.g., Assumption B-1 in Kumar et al., arXiv:2010.08585), the proposed schemes remove such a restrictive assumption. We provide a rigorous mean-square convergence analysis and establish that the proposed schemes achieve convergence rates arbitrarily close to 1/2 for the interacting particle systems associated with Levy-driven McKean-Vlasov SDEs. Several numerical examples are presented to illustrate the convergence behavior and validate the theoretical results.