Neural networks and sea time series. Recontruction and extreme-event analysis. (Q819248): Difference between revisions

This book presents the results of collaboration between a group of physicists and a group of marine engineers who have been carrying out research on a series of marine events collected from eight stations scattered around the coast of Italy and its islands. The book includes both theoretical and practical remarks concerning a long time series of sea data and the use of artificial neural networks (ANN) to solve the reconstruction problem. The first three chapters are devoted to an exposition of the main problem. The authors describe the sea events as well as the measuring techniques used in the research. Moreover, the theoretical model employed to predict sea waves and tides was also examined. Chapter 4 discusses the main probabilistic aspects of the ANN theory from the point of view of approximating a given function by means of a set of complete functions. Chapter 6 explained the methods used to analyze the probability distributions of extreme events. The numerous applications of Chapters 7, 8, 9 with their many variations and empirical insertions justify the need for a statement on the given version of the algorithms. Finally, in Chapter 10 ANN algorithms were constructed to capture the relationship of the California coast significant wave height (SWH) time series and to reconstruct gaps for missing data in the long SWH time series with the same accuracy. This chapter also sets out predictions of temperatures resulting from global warming. This book is a pioneering work. By using artificial intelligence methods the authors have built models of sea waves and tides for the Mediterranean Sea. Unfortunalely, the time series of data collection (ca. four years) is too small to be generalized from a climatological point of view. There is no comprehensive model of temperatures in conditions of global warming. Moreover, analysis of Tsunamis remains unsatisfactory. I think this research study should be continued. Nevertheless, I strongly encourage both researchers engaged in work on neural networks, data analysis, etc., as well as practitioners to study this very valuable book, which should be regarded as a ``work in progress''.