Second-order variational analysis in optimization, variational stability, and control. Theory, algorithms, applications (Q6536029)

This self-contained book, authored by a well-known expert in non-smooth analysis and optimization, is the first comprehensive work on second-order variational analysis, including both numerical algorithms and practical applications. The author has made significant contributions to the development of topics covered in this book, notably introducing the concept of second-order subdifferentials (also known as generalized Hessians) for extended-real-valued functions in the early 1990s. The work is dedicated to the author's mentors and colleagues: Y. Bogdanov, R. Gabasov, B. Polak, T. Rockafellar, and V. Tikhomirov.\N\NThe book consists of nine chapters and serves as a fundamental resource for theoretical results on the following topics:\N\N\begin{itemize}\N\item[1.] Preliminaries of first-order variational analysis, generalized differentiation, and foundational concepts in second-order analysis,\N\item[2.] Second-order subdifferential calculus,\N\item[3.] Evaluation of second-order subdifferentials,\N\item[4.] Lipschitzian stability,\N\item[5.] Full stability of local minimizers,\N\item[6.] Full stability in variational systems,\N\item[7.] Full stability in PDE optimal control,\N\item[8.] Variational convexity in optimization,\N\item[9.] Second-order numerical variational analysis.\N\end{itemize}\N\NReferencing a remarkable 350 sources, the book provides an extensive survey of theoretical results and their important applications in various fields, including machine learning, statistics, image processing and related disciplines.\N\NEach chapter includes historical context, key definitions, theorems, and examples/exercises for readers to complete. This structure makes the book a useful resource for both lecturers and students and offers inspiration for PhD students.\N\NOverall, the book is a valuable resource that can be recommended to scientific researchers, students, and anyone interested in non-smooth analysis and optimization. The author presents generalized Newtonian-type numerical algorithms in a clear and accessible manner, with practical examples that could be particularly appealing for applications in computer science and engineering. Simulations of these algorithms are conducted using MATLAB software and are compared with widely recognized algorithms published online by leading researchers. The author asserts that his algorithms are more efficient, and it would be beneficial to share these codes on a website or in repositories such as GitHub.