1 Weierstrass Institute for Applied Analysis and Stochastics, Mohrenstr. 39, D-10117 Berlin, Germany 2 Institute of Mathematics and Mechanics, Russian Academy of Sciences, S. Kovalevskaya Str. 16, 620219 Ekaterinburg, Russia
Abstract:
A number of new layer methods for solving the Neumann problemfor semilinear parabolic equations are constructed by usingprobabilistic representations of their solutions. The methodsexploit the ideas of weak-sense numerical integration of stochasticdifferential equations in a bounded domain. In spite of theprobabilistic nature these methods are nevertheless deterministic.Some convergence theorems are proved. Numerical tests on theBurgers equation are presented.