We study a non parametric estimation problem for regression models in continuous time with noises defined through non -Gaussian semi - Markov processes with jumps. Moreover, we assume that the jumps are modeled on the basis of the fractal Poisson processes with a memory in the increments. It should be noted that such models are very popular in stochastic financial markets and signal processing models. For such models we use the adaptive model selection methods developed in [2]. A sharp non-asymptotic oracle inequality for the robust risks is obtained and robust efficiency property for the Sobolev functional class is established.
Международная научная конференция "Робастная статистика и финансовая математика – 2020" (15-16 декабря 2020 г.) : сборник статей. Томск, 2021. С. 14-22