Remarks on stochastic automatic adjoint differentiation and financial models calibration

2019/01/14 by Goloubentsev, Dmitri, Lakshtanov, Evgeny
#Computational Finance (q-fin.CP) #FOS: Economics and business

paper · doi:10.48550/arxiv.1901.04200

Abstract

In this work, we discuss the Automatic Adjoint Differentiation (AAD) for functions of the form G=(1)/(2)∑₁^m (Ey_i-C_i)², which often appear in the calibration of stochastic models. We demonstrate that it allows a perfect SIMD\footnoteSingle Input Multiple Data parallelization and provide its relative computational cost. In addition we demonstrate that this theoretical result is in concordance with numeric experiments.

Related

• Automatic Adjoint Differentiation for special functions involving expectations
• Adjoints and Automatic (Algorithmic) Differentiation in Computational Finance
• Mini-symposium on automatic differentiation and its applications in the financial industry
• Automatic Backward Differentiation for American Monte-Carlo Algorithms\n (Conditional Expectation)
• Automating Steady and Unsteady Adjoints: Efficiently Utilizing Implicit and Algorithmic Differentiation