Algorithm for determining the volatility function in the Black-Scholes model (Q2300719): Difference between revisions

In the paper the authors consider an inverse problem of finding local volatility function, given the prices of a derivative instruments. They assume that the dynamics of an asset's price \(s\) is given by the stochastic differential equation of Black-Scholes type: \[ ds=s(r dt + \sigma(t)dW), \] where \(\sigma\) is an unknown volatility function. There are available observation of the prices of derivative financial instruments, which (according to the Feynaman-Kac formula) are given by the solution \(u(t,s)\) to the partial differential equation \[ \frac{\partial u}{\partial t} = sr\frac{\partial u}{\partial s} + \frac{s^2 \sigma ^2}{2}\frac{\partial^2 u}{\partial s^2} - ru. \] To estimate volatility function \(\sigma\) the authors propose to minimize a functional measuring the distance between observed prices of the derivative financial instruments, \(g(t)\) and the theoretical prices, obtained from the solution of the PDE: \[ \min_{\sigma} \int_0^T \left( u(t,s(t)) - g(t) \right)^2 dt. \] To this end the authors propose a numerical algorithm based on the modified gradient descent method. At the end the authors gives an example of the usage of the proposed algorithm.