Paper / 2017SMQ / 116025969905126

Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations

Arnulf Jentzen / Jiequn Han / Weinan E

deep learning / numerical methodsResource overview

Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations

Paper2017

by Arnulf Jentzen, Jiequn Han, Weinan E

Access resource

Continue on SelfMadeQuant

Suggested foundations for financial machine learning and optimisation and numerical methods. Specialist prerequisites vary by paper.

Time-Ordered Train, Validation and Test Design
Preserve time order and avoid future information in training.
Linear Regression
Establish an interpretable baseline before a flexible model.
Vectors, Matrices and Linear Systems
Read the linear algebra used to define a numerical problem.
Constrained and Convex Optimisation
Identify the objective, constraints and conditions for an optimum.

Finance-aware validation
Account for dependence and overlapping observations.
Feature importance
Examine which information contributes to predictions.
Calibration as an Inverse Problem
Understand identifiability and fitted model parameters.

From trading ideas to honest backtests
Choose a reproducible strategy, specify the trading rule, and test what remains after costs and estimation uncertainty.