An introduction to the theory of large deviations (Q5896501)

Let E be a Polish space an \(\{\mu_{\epsilon}\), \(\epsilon >0\}\) be a family of probability measures on E such that \(\mu_{\epsilon}\) converges weakly to \(\delta_{x_ o}\), as \(\epsilon\) \(\downarrow 0\). The problem is investigated how fast \(\mu_{\epsilon}(\Gamma)\to 0\) for \(\Gamma\in {\mathcal B}_ E\) and \(x_ 0\not\in {\bar \Gamma}\). The main part of the theorems proved is of the form: \[ \underline{\lim_{}\epsilon \downarrow 0}\epsilon \log \mu_{\epsilon}(G)\geq -\inf_{y\in G}I(y)\quad for\quad all\quad open\quad sets\quad G\subset E\quad and \] \[ \overline{\lim}_{\epsilon \downarrow 0}\quad \epsilon \log \mu_{\epsilon}(F)\leq -\inf_{y\in F}I(y)\quad for\quad all\quad closed\quad sets\quad F\subset E. \] Here I is the so-called rate function. The results and methods presented in the book belong to M. Shilder, R. Azencott, M. Kac, M. Donsker and S. R. S. Varadhan, D. Stroock, and other authors. The bibliography contains 6 positions.