General Freidlin-Wentzell large deviations and positive diffusions (Q553047): Difference between revisions

Let \(X^\varepsilon\) be the solution of the stochastic differential equation with values in \({\mathbb R}^+\) \[ dX^\varepsilon_t=b(X^\varepsilon_t) + \varepsilon \sigma(X^\varepsilon_t)dB_t,\;X^\varepsilon_0=x>0. \] In the paper, the authors prove Freidlin-Wentzel large deviations estimates under rather minimal assumptions (for cofficients \(b\) and \(\sigma\) that are not necessarily Lipschitz continuous and bounded; they are also allowed to depend on \(\varepsilon\)). This applies to models of interest in finance, i.e., the CIR and CEV models, which are positive diffusion processes whose diffusion coefficient is only Hölder continuous.