TY - JOUR
T1 - Multilevel Picard iterations for solving smooth semilinear parabolic heat equations
AU - Weinan, E.
AU - Hutzenthaler, Martin
AU - Jentzen, Arnulf
AU - Kruse, Thomas
N1 - Funding Information:
This work has been funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy EXC 2044 – 390685587, Mathematics Münster: Dynamics – Geometry – Structure, by the Deutsche Forschungsgemeinschaft via research grant HU 1889/6-1, and through the research grants ONR N00014-13-1-0338 and DOE DE-SC0009248.
Publisher Copyright:
© 2021, The Author(s).
PY - 2021/12
Y1 - 2021/12
N2 - We introduce a new family of numerical algorithms for approximating solutions of general high-dimensional semilinear parabolic partial differential equations at single space-time points. The algorithm is obtained through a delicate combination of the Feynman–Kac and the Bismut–Elworthy–Li formulas, and an approximate decomposition of the Picard fixed-point iteration with multilevel accuracy. The algorithm has been tested on a variety of semilinear partial differential equations that arise in physics and finance, with satisfactory results. Analytical tools needed for the analysis of such algorithms, including a semilinear Feynman–Kac formula, a new class of seminorms and their recursive inequalities, are also introduced. They allow us to prove for semilinear heat equations with gradient-independent nonlinearities that the computational complexity of the proposed algorithm is bounded by O(dε-(4+δ)) for any δ∈ (0 , ∞) under suitable assumptions, where d∈ N is the dimensionality of the problem and ε∈ (0 , ∞) is the prescribed accuracy. Moreover, the introduced class of numerical algorithms is also powerful for proving high-dimensional approximation capacities for deep neural networks.
AB - We introduce a new family of numerical algorithms for approximating solutions of general high-dimensional semilinear parabolic partial differential equations at single space-time points. The algorithm is obtained through a delicate combination of the Feynman–Kac and the Bismut–Elworthy–Li formulas, and an approximate decomposition of the Picard fixed-point iteration with multilevel accuracy. The algorithm has been tested on a variety of semilinear partial differential equations that arise in physics and finance, with satisfactory results. Analytical tools needed for the analysis of such algorithms, including a semilinear Feynman–Kac formula, a new class of seminorms and their recursive inequalities, are also introduced. They allow us to prove for semilinear heat equations with gradient-independent nonlinearities that the computational complexity of the proposed algorithm is bounded by O(dε-(4+δ)) for any δ∈ (0 , ∞) under suitable assumptions, where d∈ N is the dimensionality of the problem and ε∈ (0 , ∞) is the prescribed accuracy. Moreover, the introduced class of numerical algorithms is also powerful for proving high-dimensional approximation capacities for deep neural networks.
KW - Curse of dimensionality
KW - High-dimensional PDEs
KW - High-dimensional semilinear BSDEs
KW - Multilevel Monte Carlo method
KW - Multilevel Picard iteration
UR - http://www.scopus.com/inward/record.url?scp=85124521455&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85124521455&partnerID=8YFLogxK
U2 - 10.1007/s42985-021-00089-5
DO - 10.1007/s42985-021-00089-5
M3 - Article
AN - SCOPUS:85124521455
SN - 2662-2963
VL - 2
JO - Partial Differential Equations and Applications
JF - Partial Differential Equations and Applications
IS - 6
M1 - 80
ER -