TY - GEN

T1 - The Monte Carlo Method for a Solution of ODE System

AU - Tovstik, Tatiana Mikhailovna

AU - Ermakov, Sergei Mikhailovich

PY - 2019/9

Y1 - 2019/9

N2 - The Monte Carlo method is an effective method for a solution of large dimension problems.An elaboration of methods of solution of Cauchu problems for large systems of ODE basedon a simulation of Markov processes is very interesting, but this problem is not enoughinvestigated now.Here this problem is discussed and some peculiarities are indicated. As a rule differentialequations are replaced by equivalent integral Volterra equations, and then the well knownNeumann-Ulam system is used (see Ermakov S.M., 2009).The appearing peculiarities are discussed. First, the linearization that complicate theMarkov process construction is to be used. The case of a polynomial nonlinearity is anexclusive one, because there exist algorithms based on the approximation of nonlinearfunctions by polynomials and on branching processes simulation. In the general case theinterval [0,t] is divided by subintervals of length and the approximation is performed ineach subinterval. Two types of errors, systematic and random, appear and they are to beinvestigated.Second, in the linear case the valuetmay be arbitrary large, but in the nonlinear casethe choice of depends as on the small value of error of polynomial approximation, so onthe value of the Picard interval of the solution existence.As examples, some linear and nonlinear systems ODE by the Monte Carlo method aresolved. The simulation of branching Markov chains is used. The obtained solutions arecompared with the solutions found by the Runge-Kutta method.The error of the proposed methods is to be investigated in future, but the consideredexamples point of their perspective.The work is supported by RFBR,No17-01-00267-a.

AB - The Monte Carlo method is an effective method for a solution of large dimension problems.An elaboration of methods of solution of Cauchu problems for large systems of ODE basedon a simulation of Markov processes is very interesting, but this problem is not enoughinvestigated now.Here this problem is discussed and some peculiarities are indicated. As a rule differentialequations are replaced by equivalent integral Volterra equations, and then the well knownNeumann-Ulam system is used (see Ermakov S.M., 2009).The appearing peculiarities are discussed. First, the linearization that complicate theMarkov process construction is to be used. The case of a polynomial nonlinearity is anexclusive one, because there exist algorithms based on the approximation of nonlinearfunctions by polynomials and on branching processes simulation. In the general case theinterval [0,t] is divided by subintervals of length and the approximation is performed ineach subinterval. Two types of errors, systematic and random, appear and they are to beinvestigated.Second, in the linear case the valuetmay be arbitrary large, but in the nonlinear casethe choice of depends as on the small value of error of polynomial approximation, so onthe value of the Picard interval of the solution existence.As examples, some linear and nonlinear systems ODE by the Monte Carlo method aresolved. The simulation of branching Markov chains is used. The obtained solutions arecompared with the solutions found by the Runge-Kutta method.The error of the proposed methods is to be investigated in future, but the consideredexamples point of their perspective.The work is supported by RFBR,No17-01-00267-a.

KW - Monte carlo method

KW - ODE systems

M3 - Conference contribution

SP - 90

BT - 10th International Workshop on Simulation and Statistics

PB - Universitat Salzburg

CY - Salzburg

T2 - 10th International Workshop on Simulation and Statistics<br/>

Y2 - 2 September 2019 through 6 September 2019

ER -