Saved in:
Bibliographic Details
Main Authors: Seo, Jaemin, Lee, Surin, Lee, Jae Yong
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.14643
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916012561006592
author Seo, Jaemin
Lee, Surin
Lee, Jae Yong
author_facet Seo, Jaemin
Lee, Surin
Lee, Jae Yong
contents Deep learning methods based on backward stochastic differential equations (BSDEs) have emerged as competitive alternatives to physics-informed neural networks (PINNs) for solving high-dimensional partial differential equations (PDEs). By leveraging probabilistic representations, BSDE approaches can avoid the curse of dimensionality and often admit second-order-free training objectives that do not require explicit Hessian evaluations. It has recently been established that the commonly used Euler-Maruyama (EM) time discretization induces an intrinsic bias in BSDE training losses. While high-order schemes such as Heun can fully eliminate this bias, such schemes re-introduce second-order spatial derivatives and incur substantial computational overhead. In this work, we provide a principled analysis of EM-induced loss bias and propose an unbiased, second-order-free training framework that preserves the computational advantages of BSDE methods. Our code is available at https://github.com/seojaemin22/Un-EM-BSDE.
format Preprint
id arxiv_https___arxiv_org_abs_2605_14643
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Unbiased and Second-Order-Free Training for High-Dimensional PDEs
Seo, Jaemin
Lee, Surin
Lee, Jae Yong
Machine Learning
Numerical Analysis
Optimization and Control
65C30, 68TO7
Deep learning methods based on backward stochastic differential equations (BSDEs) have emerged as competitive alternatives to physics-informed neural networks (PINNs) for solving high-dimensional partial differential equations (PDEs). By leveraging probabilistic representations, BSDE approaches can avoid the curse of dimensionality and often admit second-order-free training objectives that do not require explicit Hessian evaluations. It has recently been established that the commonly used Euler-Maruyama (EM) time discretization induces an intrinsic bias in BSDE training losses. While high-order schemes such as Heun can fully eliminate this bias, such schemes re-introduce second-order spatial derivatives and incur substantial computational overhead. In this work, we provide a principled analysis of EM-induced loss bias and propose an unbiased, second-order-free training framework that preserves the computational advantages of BSDE methods. Our code is available at https://github.com/seojaemin22/Un-EM-BSDE.
title Unbiased and Second-Order-Free Training for High-Dimensional PDEs
topic Machine Learning
Numerical Analysis
Optimization and Control
65C30, 68TO7
url https://arxiv.org/abs/2605.14643