A course on rough paths. With an introduction to regularity structures (Q5920676): Difference between revisions

The book under review is a concise introduction to the theory of rough paths, their applications in stochastic analysis, and to recent extensions allowing to handle stochastic partial differential equations. The last three chapters give a short introduction to regularity structures. There exist already several other monographs on rough paths, such as [\textit{T. Lyons} and \textit{Z.-M. Qian}, System control and rough paths. Oxford: Clarendon Press (2002; Zbl 1029.93001); \textit{T. J. Lyons}, \textit{M. Caruana} and \textit{T. Lévy}, Differential equations driven by rough paths. Berlin: Springer (2007; Zbl 1176.60002); \textit{P. K. Friz} and \textit{N. B. Victoir}, Multidimensional stochastic processes as rough paths. Theory and applications. Cambridge: Cambridge University Press (2010; Zbl 1193.60053)], but the present book is a welcome addition to the literature as it contains several novelties compared to the above mentioned works. To start with, this is the first monograph based on Gubinelli's controlled path approach [\textit{M. Gubinelli}, J. Funct. Anal. 216, No. 1, 86--140 (2004; Zbl 1058.60037)], which recently has proved very powerful and was one of the main inspirations for the second author's \textit{regularity structures} [\textit{M. Hairer}, Invent. Math. 198, No. 2, 269--504 (2014; Zbl 1332.60093)]. Moreover, rough paths are a rapidly developing field, and many of the results presented here were simply not available in 2010, when the book by Friz and Victoir [loc.\,cit.]\ appeared. This concerns, most prominently, many of the applications of rough paths and, more generally, regularity structures to SPDEs, to which the last four chapters are devoted. Generally speaking, the basic problem of rough paths is to find a topology in which the solution \(Y\) to a controlled differential equations of the form \[ dY_t = f(Y_t) dX_t \] depends continuously on the driving signals \(X\). As observed by \textit{T. J. Lyons} [Rev. Mat. Iberoam. 14, No. 2, 215--310 (1998; Zbl 0923.34056)], if \(X\) is continuous and has finite \(p\)-variation, then it is sufficient (and in a certain sense also necessary) to control \(X\) and its iterated integrals up to order \(\lfloor p \rfloor\). Contrary to the previously existing monographs on the subject, the present book is not so much concerned with presenting this result in full generality, but concentrates mostly on the first nontrivial case, \(p \in [2,3)\), which already contains the most important example, namely, that of \(X\) being a Brownian motion. Also, \(p\)-variation regularity is replaced by Hölder continuity, which facilitates the presentation. The advantage of this approach is that the main ideas and techniques of rough paths can be derived succinctly and without too many technicalities, which allows for an easy entrance into the subject (arguably the easiest one available so far). Throughout, the authors have taken care to keep the material accessible and generally went for the most comprehensible rather than the most general statements. Even so, they have managed to cover an impressive amount of material in 240 pages, ranging from the basics of controlled rough paths and rough differential equations, to applications of rough paths in (Gaussian) stochastic analysis, and to stochastic partial differential equations. The main body of the text contains most of the fundamental results, while many further results, often from the forefront of current research, are delegated to the exercises. In fact, the plethora of exercises after each chapter (sometimes with solution, sometimes with hints, sometimes without any help) is certainly one of the highlights of the book. It is easy to base a graduate course on rough paths on this monograph. In principle, Chapters 13--15 could also build the skeleton for a course on regularity structures, albeit, since these chapters contain essentially no proofs (notable exception: the reconstruction theorem), one would need another reference to fill in the details. On the other side, a researcher who carefully works her way through all of the exercises will have a very good impression of the current state of the art in rough paths (with the exception of some omissions that are not covered in the text, such as signatures). We continue with an overview of the book: Chapter 1 is the introduction. Chapter 2 and 4 cover the basics of rough paths, controlled paths, and the controlled path integral. The main focus is on \(\alpha\)-Hölder continuous rough paths with \(\alpha > 1/3\), but it is indicated how to adapt the arguments in order to treat rougher paths. In particular, the integral is constructed with the help of the \textit{sewing lemma} (Lemma 4.2), which extends without a problem to higher order rough paths. To keep the reader motivated, the guiding example of the Brownian rough path is introduced right away in Chapter 3. Chapter 5 links rough path integration with respect to the Brownian rough path to stochastic integration (Itō, Stratonovich, backward Itō). It also builds a bridge between rough path integration and Föllmer's pathwise Itō calculus. Many of the results in Chapter 5 are new. Chapter 6 is concerned with the question under which conditions the Gubinelli derivative is unique. If \(X\) is an \(\alpha\)-Hölder continuous path, then \(Y\) is called controlled by \(X\) if there exists an \(\alpha\)-Hölder continuous \(Y'\) such that \[ Y_t - Y_s = Y'_s (X_t - X_s) + O(|t-s|^{2\alpha}). \] In this case, \(Y'\) is called the Gubinelli derivative. Clearly, it is not unique in general, for example, if \(Y\) and \(X\) are both \(2\alpha\)-Hölder continuous, then \(Y'\) can be chosen arbitrarily in \(C^\alpha\). Here, it is shown that under suitable ``roughness'' conditions on \(X\), the Gubinelli derivative is indeed unique. Chapter 7 sets the stage for rough differential equations by studying the structure of the space of controlled paths and its stability under the application of nonlinear maps or under multiplication. Chapter~8 then deals with rough differential equations and derives the basic existence and uniqueness results, as well as a priori estimates on the solutions, and the continuity of the Itō-Lyons map in rough path topology. These results are presented in the controlled path approach, but also Lyons' original approach and the one of \textit{A. M. Davie} [AMRX, Appl. Math. Res. Express 2007, Article ID abm009, 40 p. (2008; Zbl 1163.34005)] are briefly discussed. Chapter~9 presents some applications to Brownian SDEs, such as a support theorem and a large deviation result. Chapters 10 and 11 go beyond the Brownian case and deal with SDEs driven by irregular Gaussian signals, such as the fractional Brownian motion with Hurst index \(H \in (1/3,1/2)\). In that case, it is possible to obtain good tail estimates for the solutions, as well as a version of Hörmander's theorem in a non-Markovian setting. Chapter 12 contains some recent applications of rough paths to SPDEs. The first part of the Chapter deals with equations driven by finite dimensional signals, for which several approaches such as flow transformations, viscosity solutions, and Feynman-Kac representations are discussed. The second part of this chapter is concerned with the SPDE \[ \partial_t u = \partial^2_{x} u + g(u) \partial_x u + \xi, \] where \(\xi\) is a space-time white noise. This equation was first solved by \textit{M. Hairer} [Commun. Pure Appl. Math. 64, No. 11, 1547--1585 (2011; Zbl 1229.60079)], who realized that the controlled rough path integral allows us to make sense of the nonlinearity \(g(u) \partial_x u\) and then to solve the equation. This insight was a major breakthrough that ultimately led to the development of regularity structures. Here, the rough path approach is only sketched and the equation is not completely solved, but it is carefully shown how to construct the underlying rough path on which the new solution theory is based. Chapters 13--15 form a brief introduction to regularity structures and can be essentially read independently of the rest, although the space of controlled rough paths serves as a recurring example. There are essentially no proofs, so this part should be understood as a teaser, inviting the reader to have a look at Hairer's original paper. Chapter 13 sets up the basic definitions and properties of regularity structures, while Chapter 14 studies the space of modelled distributions. Chapter 15 sketches how to apply regularity structures in order to solve the KPZ equation.