Adjoint-based Monte Carlo calibration of financial methods (Q964678)

Adjoint methods allow to quickly compute option sensitivities with respect to a large number of model parameters. The authors illustrate for the case of calibrating a general \(L\)-dimensional diffusion model to a given set of call option prices that it is possible to successfully implement an adjoint-based Monte-Carlo algorithm. After discretizing the problem in the sense of a sample average approximation and smoothing out potential nondifferentiabilities they obtain an approximation of the original calibration problem with continuously differentiable coefficient functions. Then the convergence results for the solutions of subproblems are obtained. Some algorithmic and computational issues regarding the implementation of the optimization algorithm are discussed and the derivation of the adjoint equation in the discretized diffusion model is studied.